Say which bytes the byte limits count, and what passing one does (closes #203)

The byte limits and the byte anomaly thresholds count body bytes, as
SWW_BYTES_COUNT chooses them, once a response has ended; the request
that takes a client past a limit is passed on whole, and the client is
banned. README.md, SPEC.md and docs/settings.md now say so where they
said "bytes a client's requests may carry". Also made plain: a request
past a rate limit is refused and bans the client; SWRP_MAX_REQUEST_TIME_MS
is two consecutive limits; SWW_LOOKUP_SOURCE's three values; and the
claim that discriminator always answers within SWRP_TIMEOUT_MS, which
the spec itself contradicts.

Model: fable-5-1

Dockerfile

@@ -136,10 +136,19 @@ RUN mkdir /etc/sysusers.d \
 # ca-certificates, nix-bin and runit, from Ubuntu's archive. apt fetches
 # them over plain HTTP and checks each one against the archive's signed
 # package lists, so no CA certificates are needed.
-RUN apt-get update \
+#
+# nix-bin depends on xz-utils, whose install links man pages with
+# update-alternatives. The image's dpkg config drops man pages, so each
+# link warns. z-man, read after /etc/dpkg/dpkg.cfg.d/excludes, re-includes
+# the man1 pages for this install alone; the same step deletes them again,
+# so no man page or package is added.
+RUN echo 'path-include=/usr/share/man/man1/*' > /etc/dpkg/dpkg.cfg.d/z-man \
+    && apt-get update \
     && DEBIAN_FRONTEND=noninteractive apt-get install --yes --no-install-recommends \
         ca-certificates nix-bin runit \
-    && rm -rf /var/lib/apt/lists/*
+    && rm -rf /var/lib/apt/lists/* /etc/dpkg/dpkg.cfg.d/z-man \
+    && find /usr/share/man/man1 ! -type d -delete \
+    && find /etc/alternatives -type l -xtype l -delete
 
 # Nix run by root expects a group of build users, which nix-bin does not
 # create; with the setting empty, root's builds run without them.

README.md

@@ -25,11 +25,11 @@ without an answer in time, `rproxy` passes the request to the app.
 It protects against:
 
 - Scrapers that fetch every page of a site, all day. Each client gets a limit on
-  requests and on bytes per minute, per hour and per day. The defaults leave
+  requests and on body bytes per minute, per hour and per day. The defaults
-  room for a person reading a site for hours. They catch a scraper that keeps up
+  leave room for a person reading a site for hours. They catch a scraper that
-  one request every 2 seconds. A client over a limit is refused and banned: for
+  keeps up one request every 2 seconds. A client that breaks a limit is banned:
-  a while at first, for three times as long at each repeat, and permanently when
+  for a while at first, for three times as long at each repeat, and permanently
-  it keeps at it.
+  when it keeps at it.
 - Scanners that probe for secrets, such as `/.env` or `/.git/config`. The
   default rule file bans such a client at its first probe, and its next request
   during the ban makes the ban permanent. Your own rules, or a list of trap
@@ -209,6 +209,9 @@ account, a key, a token or an alert destination stays off until you give one.
   percentage is a whole number from 0 to 100. A path prefix starts with `/`.
 - `off` switches a limit, a threshold or a timeout off where the table says a
   setting takes it.
+- The byte limits and the anomaly thresholds on bytes count body bytes: the
+  app's response bodies, the visitors' request bodies, or both, as
+  `SWW_BYTES_COUNT` says. Headers are never counted.
 - An invalid setting stops the process at start with one log line naming the
   variable and what is wrong with the value. The settings in effect are logged
   on one line at start, with tokens, keys and webhook paths shown as `********`.
@@ -216,7 +219,7 @@ account, a key, a token or an alert destination stays off until you give one.
 Every setting the code reads, with its default:
 
 | Variable                                 | Default                                                                        | What it does                                                                                                                                                                                            |
-| ---------------------------------------- | ------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| ---------------------------------------- | ------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
 | `SWW_COMMAND`                            | `rproxy`                                                                       | Which command the image runs, and its health check checks: `rproxy` or `discriminator`.                                                                                                                 |
 | `SWRP_WAF_BACKEND`                       | required                                                                       | `discriminator`'s URL, such as `http://discriminator:8080`.                                                                                                                                             |
 | `SWRP_UPSTREAM_URL`                      | `http://127.0.0.1:8081`                                                        | The app's URL: a scheme, a host and an optional port, nothing more.                                                                                                                                     |
@@ -226,7 +229,7 @@ Every setting the code reads, with its default:
 | `SWRP_DEFAULT_DISPOSITION`               | `allow`                                                                        | What `rproxy` does with a request when `discriminator` cannot be reached or does not answer in time: `allow` passes it to the app, `deny` answers `503`.                                                |
 | `SWRP_MAX_REQUEST_SIZE`                  | `100M`                                                                         | The largest request body, a size or `off`; a larger one is answered `413`.                                                                                                                              |
 | `SWRP_MAX_RESPONSE_SIZE`                 | `1G`                                                                           | The largest response body, a size or `off`; a larger one is answered `502`, or cut off once it has started.                                                                                             |
-| `SWRP_MAX_REQUEST_TIME_MS`               | `60000`                                                                        | How long the visitor may take to send its request line and headers, and then the whole request to reach the app, in milliseconds or `off`.                        |
+| `SWRP_MAX_REQUEST_TIME_MS`               | `60000`                                                                        | Two time limits, each this long, in milliseconds or `off`: one for the visitor to send its request line and headers, then one for the whole request to reach the app.                                   |
 | `SWRP_MAX_RESPONSE_TIME_MS`              | `1800000`                                                                      | How long the answer may take, from the end of the request to its last byte reaching the visitor, in milliseconds or `off`.                                                                              |
 | `SWW_LISTEN_ADDR`                        | `:8080`                                                                        | Where `discriminator` listens; leave the host part empty.                                                                                                                                               |
 | `SWW_INSTANCE_NAME`                      | the container's host name                                                      | The name in every line `discriminator` logs, and in every alert and metric, such as `fsn1/discriminator`.                                                                                               |
@@ -237,22 +240,22 @@ Every setting the code reads, with its default:
 | `SWW_ALLOW_NETS`                         | empty                                                                          | Netblocks whose clients skip every check, such as your monitoring.                                                                                                                                      |
 | `SWW_RATE_LIMIT_EXEMPT_NETS`             | empty                                                                          | Netblocks whose clients the request and byte limits neither count nor refuse.                                                                                                                           |
 | `SWW_DENY_NETS`                          | empty                                                                          | Netblocks whose clients are refused.                                                                                                                                                                    |
-| `SWW_RATE_LIMIT_PER_MINUTE`              | `120`                                                                          | The most requests a client may make in a minute, a count or `off`; one more bans it.                                                                              |
+| `SWW_RATE_LIMIT_PER_MINUTE`              | `120`                                                                          | The most requests a client may make in a minute, a count or `off`; the next request is refused and bans the client.                                                                                     |
-| `SWW_RATE_LIMIT_PER_HOUR`                | `1500`                                                                         | The most requests a client may make in an hour, a count or `off`; one more bans it.                                                                               |
+| `SWW_RATE_LIMIT_PER_HOUR`                | `1500`                                                                         | The most requests a client may make in an hour, a count or `off`; the next request is refused and bans the client.                                                                                      |
-| `SWW_RATE_LIMIT_PER_DAY`                 | `6000`                                                                         | The most requests a client may make in a day, a count or `off`; one more bans it.                                                                                 |
+| `SWW_RATE_LIMIT_PER_DAY`                 | `6000`                                                                         | The most requests a client may make in a day, a count or `off`; the next request is refused and bans the client.                                                                                        |
 | `SWW_RATE_LIMIT_EXEMPT_PATHS`            | empty                                                                          | Path prefixes whose requests the request and byte limits neither count nor refuse, such as `/assets/`.                                                                                                  |
-| `SWW_BYTES_LIMIT_PER_MINUTE`             | `off`                                                                          | The most bytes a client's requests may carry in a minute, a size or `off`; the request that takes a client past any byte limit bans it.                           |
+| `SWW_BYTES_LIMIT_PER_MINUTE`             | `off`                                                                          | The most body bytes a client's requests may total in a minute, a size or `off`; the request that takes the total past a byte limit is passed on whole, and bans the client once its response has ended. |
-| `SWW_BYTES_LIMIT_PER_HOUR`               | `50G`                                                                          | The most bytes a client's requests may carry in an hour, a size or `off`.                                                                                         |
+| `SWW_BYTES_LIMIT_PER_HOUR`               | `50G`                                                                          | The most body bytes a client's requests may total in an hour, a size or `off`.                                                                                                                          |
-| `SWW_BYTES_LIMIT_PER_DAY`                | `100G`                                                                         | The most bytes a client's requests may carry in a day, a size or `off`.                                                                                           |
+| `SWW_BYTES_LIMIT_PER_DAY`                | `100G`                                                                         | The most body bytes a client's requests may total in a day, a size or `off`.                                                                                                                            |
-| `SWW_BYTES_COUNT`                        | `both`                                                                         | Which body bytes the byte limits count: `response`, `request` or `both`.                                                                                          |
+| `SWW_BYTES_COUNT`                        | `both`                                                                         | Which body bytes the byte limits and the anomaly thresholds on bytes count: `response`, the app's response bodies; `request`, the visitors' request bodies; or `both`.                                  |
-| `SWW_ERROR_BURST_THRESHOLD`              | `30`                                                                           | The most refusals by a rule file, a trap path or the Core Rule Set a client may get in a minute, a count or `off`; one more bans it.                              |
+| `SWW_ERROR_BURST_THRESHOLD`              | `30`                                                                           | The most refusals by a rule file, a trap path or the Core Rule Set a client may get in a minute, a count or `off`; the next such refusal bans the client.                                               |
 | `SWW_BAN_RESPONSE`                       | `403`                                                                          | How a refused client is answered: `403`, `429` or `close`, which drops the connection.                                                                                                                  |
 | `SWW_LIMIT_BAN_DURATION`                 | `1h`                                                                           | The ban for a first broken limit.                                                                                                                                                                       |
 | `SWW_LIMIT_BAN_REPEAT_WINDOW`            | `24h`                                                                          | A limit broken again within this time after the last such ban ended gets a ban three times as long.                                                                                                     |
 | `SWW_MAX_BAN_DURATION`                   | `7d`                                                                           | A ban for a broken limit that would be longer than this is permanent instead.                                                                                                                           |
 | `SWW_ATTACK_BAN_DURATION`                | `7d`                                                                           | The ban for a first clear sign of attack: a `ban` rule or a trap path.                                                                                                                                  |
 | `SWW_BAN_SCOPE_V4_PREFIX`                | `32`                                                                           | The prefix length of the netblock a ban on an IPv4 client covers, from 0 to 32.                                                                                                                         |
-| `SWW_LOOKUP_SOURCE`                      | `geojs`                                                                        | Where each client's AS number and country come from: `geojs`, a web service, `file`, or `off`.                                                                    |
+| `SWW_LOOKUP_SOURCE`                      | `geojs`                                                                        | Where each client's AS number and country come from: `geojs` (a web service), `file` (a local database) or `off`.                                                                                       |
 | `SWW_LOOKUP_DB_PATH`                     | empty                                                                          | The IPinfo Lite database file, in its `.mmdb` form, read while the source is `file`.                                                                                                                    |
 | `SWW_LOOKUP_TIMEOUT`                     | `800ms`                                                                        | How long a request waits for its client's first GeoJS answer when a setting needs it, and how long a request to GeoJS may take.                                                                         |
 | `SWW_ADD_LOOKUP_HEADERS`                 | `false`                                                                        | `true` passes the app the client's AS number and country in `X-Client-ASN` and `X-Client-Country`.                                                                                                      |
@@ -310,27 +313,27 @@ Every setting the code reads, with its default:
 | `SWW_ALERT_MAX_PER_HOUR`                 | `60`                                                                           | The most alerts sent in an hour, a number or `off`; the rest are counted in the hour's `summary`.                                                                                                       |
 | `SWW_ANOMALY_CLIENT_REQUESTS_PER_MINUTE` | `off`                                                                          | Alert when one client makes more requests than this in a minute, a count.                                                                                                                               |
 | `SWW_ANOMALY_CLIENT_REQUESTS_PER_HOUR`   | `off`                                                                          | Alert when one client makes more requests than this in an hour, a count.                                                                                                                                |
-| `SWW_ANOMALY_CLIENT_BYTES_PER_MINUTE`    | `off`                                                                          | Alert when one client's requests carry more bytes than this in a minute, a size.                                                                                  |
+| `SWW_ANOMALY_CLIENT_BYTES_PER_MINUTE`    | `off`                                                                          | Alert when one client's body bytes total more than this in a minute, a size.                                                                                                                            |
-| `SWW_ANOMALY_CLIENT_BYTES_PER_HOUR`      | `off`                                                                          | Alert when one client's requests carry more bytes than this in an hour, a size.                                                                                   |
+| `SWW_ANOMALY_CLIENT_BYTES_PER_HOUR`      | `off`                                                                          | Alert when one client's body bytes total more than this in an hour, a size.                                                                                                                             |
 | `SWW_ANOMALY_NET_REQUESTS_PER_MINUTE`    | `off`                                                                          | Alert when the netblock around a client makes more requests than this in a minute, a count.                                                                                                             |
 | `SWW_ANOMALY_NET_REQUESTS_PER_HOUR`      | `off`                                                                          | Alert when the netblock around a client makes more requests than this in an hour, a count.                                                                                                              |
-| `SWW_ANOMALY_NET_BYTES_PER_MINUTE`       | `off`                                                                          | Alert when the netblock around a client carries more bytes than this in a minute, a size.                                                                         |
+| `SWW_ANOMALY_NET_BYTES_PER_MINUTE`       | `off`                                                                          | Alert when the body bytes of the netblock around a client total more than this in a minute, a size.                                                                                                     |
-| `SWW_ANOMALY_NET_BYTES_PER_HOUR`         | `off`                                                                          | Alert when the netblock around a client carries more bytes than this in an hour, a size.                                                                          |
+| `SWW_ANOMALY_NET_BYTES_PER_HOUR`         | `off`                                                                          | Alert when the body bytes of the netblock around a client total more than this in an hour, a size.                                                                                                      |
 | `SWW_ANOMALY_NET_V4_PREFIX`              | `24`                                                                           | The prefix length of the netblock around an IPv4 client that the `SWW_ANOMALY_NET_` thresholds count, from 0 to 32.                                                                                     |
 | `SWW_ANOMALY_NET_V6_PREFIX`              | `48`                                                                           | The prefix length of the netblock around an IPv6 client that the `SWW_ANOMALY_NET_` thresholds count, from 0 to 128.                                                                                    |
 | `SWW_ANOMALY_ASN_REQUESTS_PER_MINUTE`    | `off`                                                                          | Alert when one AS number makes more requests than this in a minute, a count.                                                                                                                            |
 | `SWW_ANOMALY_ASN_REQUESTS_PER_HOUR`      | `off`                                                                          | Alert when one AS number makes more requests than this in an hour, a count.                                                                                                                             |
-| `SWW_ANOMALY_ASN_BYTES_PER_MINUTE`       | `off`                                                                          | Alert when one AS number carries more bytes than this in a minute, a size.                                                                                        |
+| `SWW_ANOMALY_ASN_BYTES_PER_MINUTE`       | `off`                                                                          | Alert when the body bytes of one AS number total more than this in a minute, a size.                                                                                                                    |
-| `SWW_ANOMALY_ASN_BYTES_PER_HOUR`         | `off`                                                                          | Alert when one AS number carries more bytes than this in an hour, a size.                                                                                         |
+| `SWW_ANOMALY_ASN_BYTES_PER_HOUR`         | `off`                                                                          | Alert when the body bytes of one AS number total more than this in an hour, a size.                                                                                                                     |
 | `SWW_ANOMALY_TOTAL_REQUESTS_PER_MINUTE`  | `off`                                                                          | Alert when the whole service gets more requests than this in a minute, a count.                                                                                                                         |
 | `SWW_ANOMALY_TOTAL_REQUESTS_PER_HOUR`    | `off`                                                                          | Alert when the whole service gets more requests than this in an hour, a count.                                                                                                                          |
-| `SWW_ANOMALY_TOTAL_BYTES_PER_MINUTE`     | `off`                                                                          | Alert when the whole service carries more bytes than this in a minute, a size.                                                                                    |
+| `SWW_ANOMALY_TOTAL_BYTES_PER_MINUTE`     | `off`                                                                          | Alert when the body bytes of the whole service total more than this in a minute, a size.                                                                                                                |
-| `SWW_ANOMALY_TOTAL_BYTES_PER_HOUR`       | `off`                                                                          | Alert when the whole service carries more bytes than this in an hour, a size.                                                                                     |
+| `SWW_ANOMALY_TOTAL_BYTES_PER_HOUR`       | `off`                                                                          | Alert when the body bytes of the whole service total more than this in an hour, a size.                                                                                                                 |
 | `SWW_WATCH_NETS`                         | empty                                                                          | Named netblocks the `SWW_WATCH_` thresholds count each as a whole, such as `office=203.0.113.0/24,scraper-x=198.51.100.0/22`.                                                                           |
 | `SWW_WATCH_REQUESTS_PER_MINUTE`          | `off`                                                                          | Alert when a watched netblock makes more requests than this in a minute, a count.                                                                                                                       |
 | `SWW_WATCH_REQUESTS_PER_HOUR`            | `off`                                                                          | Alert when a watched netblock makes more requests than this in an hour, a count.                                                                                                                        |
-| `SWW_WATCH_BYTES_PER_MINUTE`             | `off`                                                                          | Alert when a watched netblock carries more bytes than this in a minute, a size.                                                                                   |
+| `SWW_WATCH_BYTES_PER_MINUTE`             | `off`                                                                          | Alert when the body bytes of a watched netblock total more than this in a minute, a size.                                                                                                               |
-| `SWW_WATCH_BYTES_PER_HOUR`               | `off`                                                                          | Alert when a watched netblock carries more bytes than this in an hour, a size.                                                                                    |
+| `SWW_WATCH_BYTES_PER_HOUR`               | `off`                                                                          | Alert when the body bytes of a watched netblock total more than this in an hour, a size.                                                                                                                |
 
 [`docs/settings.md`](docs/settings.md) gives the values each setting takes in
 full, and the limits that are fixed rather than settings.
@@ -524,8 +527,9 @@ of the request reaches the app.
 
 14. The request is passed to the app, streaming.
 15. After the answer, `rproxy` sends `discriminator` a report with the status
-    and the bytes each way. For an allowed request, `discriminator` counts the
+    and the body bytes each way. For an allowed request, `discriminator` counts
-    bytes toward the byte limits and adds the request to the client's history.
+    the bytes `SWW_BYTES_COUNT` names toward the byte limits, bans the client if
+    they take it past one, and adds the request to the client's history.
 
 A refused visitor is answered with `SWW_BAN_RESPONSE`. A `block` rule or a Core
 Rule Set match answers `403` and bans nobody, since the Core Rule Set has false

SPEC.md

@@ -124,7 +124,9 @@ to them.
 step that produces a final answer ends the checks. Step 15 runs for each request
 `discriminator` answers with `allow`, `deny` or `close`. `rproxy`'s size and
 time limits apply to the whole exchange. Each step names the section that holds
-its rules.
+its rules. A step that names a heuristic in parentheses runs that heuristic. The
+request log's `heuristic` field gives that name. `refusal_limit` and
+`byte_limits` give theirs in `after_answer_heuristic` instead.
 
 1. `GET /_sww/healthz` is answered at once by `rproxy`, before any check, so a
    health checker never needs an exemption and is never refused.
@@ -142,9 +144,9 @@ its rules.
       around its address, a /64 by default, since one abuser usually controls a
       whole /64. A client is therefore one IPv4 address or one IPv6 group.
 3. The static lists. A client in `SWW_ALLOW_NETS` skips steps 4 to 12, the byte
-   limits and the error burst. A client in `SWW_DENY_NETS` is refused; a client
+   limits and the error burst. A client in `SWW_DENY_NETS` is refused
-   in both lists is let through.
+   (`deny_nets`); a client in both lists is let through.
-4. Bans. A client under a ban is refused (see "Bans").
+4. Bans. A client under a ban is refused (`active_ban`; see "Bans").
 5. The lookup, unless `SWW_LOOKUP_SOURCE` is `off` (see "Lookup of AS number and
    country"). The request goes on at once. It waits for the first GeoJS answer,
    up to `SWW_LOOKUP_TIMEOUT`, only when one of these needs it:
@@ -154,21 +156,25 @@ its rules.
 
     A request that ends before the answer comes is logged without it.
 
-6. The country lists (see "Country lists").
+6. The country lists (`countries`; see "Country lists").
-7. The reputation sources (see "Reputation"). A client on a fetched blocklist
+7. The reputation sources (see "Reputation"), in this order:
-   gets `SWW_BLOCKLIST_ACTION`. A client on the CrowdSec decision list is
+    - A client on a fetched blocklist gets `SWW_BLOCKLIST_ACTION`
-   banned. A client listed by a DNS blocklist, or scored at least
+      (`blocklists`).
-   `SWW_ABUSEIPDB_MIN_SCORE` by AbuseIPDB, gets `SWW_REPUTATION_ACTION`.
+    - A client on the CrowdSec decision list is banned (`crowdsec`).
+    - A client listed by a DNS blocklist gets `SWW_REPUTATION_ACTION` (`dnsbl`).
+    - A client scored at least `SWW_ABUSEIPDB_MIN_SCORE` by AbuseIPDB gets
+      `SWW_REPUTATION_ACTION` (`abuseipdb`).
 8. The client's limit percentage (see "Biased thresholds").
-9. Rate limiting (see "Request rate limits, per client"). A request over a limit
+9. Rate limiting (`rate_limits`; see "Request rate limits, per client"). A
-   times the percentage is refused and bans the client.
+   request over a limit times the percentage is refused and bans the client.
-10. The trap paths (see "Attack detection"). A request for one is refused and
+10. The trap paths (`trap_paths`; see "Attack detection"). A request for one is
-    bans the client.
+    refused and bans the client.
-11. The rule files (see "Rule files"). Matching stops at the first rule that
+11. The rule files (`rule_files`; see "Rule files"). Matching stops at the first
-    refuses.
+    rule that refuses.
-12. The header limits (see "Answering in time"), and then the Core Rule Set (see
+12. The header limits (`header_limits`; see "Answering in time"), and then the
-    "Attack detection"). A Core Rule Set match is refused with `403` while
+    Core Rule Set (`core_rule_set`; see "Attack detection"). A Core Rule Set
-    `SWW_WAF_MODE` is `block`, and only logged and alerted while it is `detect`.
+    match is refused with `403` while `SWW_WAF_MODE` is `block`, and only logged
+    and alerted while it is `detect`.
 13. A path under `/_sww/` is answered `404`, so that the app never sees that
     prefix.
 14. `rproxy` passes the request to `SWRP_UPSTREAM_URL`, streaming (see "Passing
@@ -178,18 +184,19 @@ its rules.
     sent is removed.
 15. After the visitor's answer has ended, `rproxy` sends the report (see "The
     report"). `discriminator` then counts:
-    - For a request it allowed, the response bytes, the request bytes or both
+    - For a request it allowed, the body bytes of the response, of the request
-      are added to the byte counters, as `SWW_BYTES_COUNT` says. For a
+      or of both are added to the byte counters, as `SWW_BYTES_COUNT` says. For
-      WebSocket, what it carried each way counts once it closes. A client whose
+      a WebSocket, what it carried each way counts once it closes. A client
-      total passes a byte limit times the percentage has broken that limit and
+      whose total passes a byte limit times the percentage has broken that limit
-      is banned. The response in progress is not cut off.
+      and is banned (`byte_limits`). The request that broke it has already been
+      passed on whole; nothing is cut off.
     - For a request it allowed, the client's history and the anomaly counters
       are updated, and the alert thresholds are evaluated.
 
     For a request it refused, `discriminator` adds nothing to the byte limits.
     It updates the history and the anomaly counters at its decision. It counts
-    the refusals the error burst counts at the decision too (see "Attack
+    the refusals the error burst counts at the decision too (`refusal_limit`;
-    detection").
+    see "Attack detection").
 
 In `SWW_MODE=observe`, every refusal above is logged as what would have
 happened, nothing is refused, and `discriminator` makes no ban of its own. A
@@ -245,7 +252,7 @@ to a ban, each by its own rule:
       clear sign of attack bans it permanently at once.
 - A broken limit, one of:
     - a request over a rate limit;
-    - a response that takes the client's byte total past a byte limit;
+    - a request whose body bytes take the client's total past a byte limit;
     - more than `SWW_ERROR_BURST_THRESHOLD` refusals within a minute.
 
     The bans for a broken limit:
@@ -433,10 +440,13 @@ the same way.
 
 ### Byte limits, per client
 
-By default there is no minute limit, so a client can download in a burst at full
+The byte limits count body bytes: the app's response bodies, the visitors'
-speed. The hour limit defaults to 50 GiB and the day limit to 100 GiB. They
+request bodies, or both, as `SWW_BYTES_COUNT` says. Headers are never counted. A
-allow several downloads of a few GiB each. A response's bytes are counted when
+request's bytes are counted once its response has ended (see step 15). So the
-it ends (see step 15).
+request that takes a client past a limit is passed on whole, and the ban refuses
+the client's requests from then on. By default there is no minute limit, so a
+client can download in a burst at full speed. The hour limit defaults to 50 GiB
+and the day limit to 100 GiB. They allow several downloads of a few GiB each.
 
 ### Lookup of AS number and country
 
@@ -899,7 +909,8 @@ which the metrics show. They reveal a widely distributed scraper, thousands of
 addresses at low rates each, which the per-client limits do not see. A low
 `SWW_ASN_LIMIT_PERCENT` for its AS number is the response.
 
-- Requests and bytes are counted per minute and per hour, for each of:
+- Requests and body bytes, as `SWW_BYTES_COUNT` counts them, are counted per
+  minute and per hour, for each of:
     - a client;
     - the surrounding netblock;
     - an AS number;
@@ -1209,9 +1220,9 @@ short.
 ### Answering in time
 
 `rproxy` stops waiting for `discriminator` after `SWRP_TIMEOUT_MS`, by default 2
-seconds, and `SWRP_DEFAULT_DISPOSITION` then decides. So at the defaults,
+seconds, and `SWRP_DEFAULT_DISPOSITION` then decides. So the longest a question
-`discriminator` answers a request well within that time, whatever the request
+waits on an outside source, `SWW_LOOKUP_TIMEOUT`, is by default well within that
-and whatever its outside sources do.
+time, whatever the source does.
 
 GeoJS is the only outside source a question can wait on. A question waits only
 when step 5 says so, and for at most `SWW_LOOKUP_TIMEOUT`, by default 800
@@ -1307,9 +1318,9 @@ answer.
 For a request it holds, `discriminator` writes the request's last log line (see
 "The log"). For a request it allowed, it first:
 
-- counts the bytes as `SWW_BYTES_COUNT` says, toward the client's byte limits at
+- counts the body bytes `SWW_BYTES_COUNT` names, toward the client's byte limits
-  the percentages it gave that request;
+  at the percentages it gave that request;
-- bans a client that passes a byte limit;
+- bans a client whose total passes a byte limit;
 - adds the request with its status to the client's history, the anomaly counters
   and the metrics.
 

docs/request-log.md

@@ -108,6 +108,35 @@ Every one of these lines has:
 - `would_action`: in `observe` mode, what `enforce` mode would have done with a
   request it would have refused. `action` then names what was done, normally
   `forward`.
+- `heuristic`: the heuristic that refused the request, in `observe` mode too. It
+  is left out when none did.
+- `heuristic_figures`: an object of the figures that heuristic acted on. Each
+  heuristic and its figures:
+    - `deny_nets`: `netblock`, the entry of `SWW_DENY_NETS` the client is in;
+    - `active_ban`: `netblock`, the ban's netblock, and `ban_expires`;
+    - `countries`: `country`, and `setting`, the country list that refused it;
+    - `blocklists`: `lists`, the blocklist URLs that list the client;
+    - `crowdsec`: `scenario`, the decision's scenario, and `expires`, when the
+      decision ends;
+    - `dnsbl`: `zones`, the DNS blocklist zones that list the client, their keys
+      masked;
+    - `abuseipdb`: `score`, and `min_score`, `SWW_ABUSEIPDB_MIN_SCORE`;
+    - `rate_limits`: `window`, `count` and `limit`;
+    - `trap_paths`: `path`, the trap path asked for;
+    - `rule_files`: `rule_id` and `rule_action`, of the rule that refused it;
+    - `header_limits`: `fields`, `fields_limit`, `cookies` and `cookies_limit`;
+    - `core_rule_set`: `score`, and `threshold`, `SWW_WAF_ANOMALY_THRESHOLD`;
+    - `refusal_limit`: `window`, `count` and `limit`, of the error burst;
+    - `byte_limits`: `window`, `count` and `limit`, the count in bytes.
+- `after_answer_heuristic`: the heuristic that banned the client after
+  `discriminator` answered the request, in `observe` mode too: `refusal_limit`
+  or `byte_limits`.
+- `after_answer_heuristic_figures`: an object of the figures that heuristic
+  acted on, as listed above.
+
+    `byte_limits` bans when the report comes, so only the `report` line names
+    it.
+
 - `limit_percent` and `limit_percent_setting`: the client's percentage of the
   rate limits and the setting that lowered it, when one did. `bytes_percent` and
   `bytes_percent_setting` do the same for the byte limits.

docs/settings.md

@@ -98,6 +98,8 @@ between the two (see "The image's command" below).
 
 ## Request rate limits, per client
 
+The request past one of these limits is refused, and the client is banned.
+
 - `SWW_RATE_LIMIT_PER_MINUTE` (default `120`, 2 a second): the most requests a
   client may make in a minute. A count or `off`. The default leaves room for a
   person opening two pages at once with an empty cache, up to 60 files each.
@@ -120,11 +122,16 @@ between the two (see "The image's command" below).
 
 ## Byte limits, per client
 
+These count body bytes, never headers. A request's bytes are counted once its
+response has ended. The request that takes a client's total past a limit is not
+cut off; the client is banned.
+
 - `SWW_BYTES_LIMIT_PER_MINUTE` (default `off`), `SWW_BYTES_LIMIT_PER_HOUR`
-  (default `50G`) and `SWW_BYTES_LIMIT_PER_DAY` (default `100G`): the most bytes
+  (default `50G`) and `SWW_BYTES_LIMIT_PER_DAY` (default `100G`): the most body
-  counted for a client in each window. A size or `off`.
+  bytes a client's requests may total in each window. A size or `off`.
-- `SWW_BYTES_COUNT` (default `both`): which body bytes the byte limits count:
+- `SWW_BYTES_COUNT` (default `both`): which body bytes the byte limits and the
-  `response`, `request` or `both`.
+  anomaly thresholds on bytes count: `response`, the app's response bodies;
+  `request`, the visitors' request bodies; or `both`.
 
 ## Lookup of AS number and country
 
@@ -276,9 +283,10 @@ between the two (see "The image's command" below).
 
 ## Anomaly thresholds
 
-Each scope has four thresholds: the most requests and the most bytes counted in
+Each scope has four thresholds: the most requests and the most body bytes
-a minute and in an hour. Requests are counts, bytes are sizes. Each is `off` by
+counted in a minute and in an hour. The bytes are those `SWW_BYTES_COUNT` names.
-default. The four thresholds per client are:
+Requests are counts, bytes are sizes. Each is `off` by default. The four
+thresholds per client are:
 
 - `SWW_ANOMALY_CLIENT_REQUESTS_PER_MINUTE`;
 - `SWW_ANOMALY_CLIENT_REQUESTS_PER_HOUR`;

docs/state-files.md

@@ -61,7 +61,7 @@ since it was first seen:
 - the first and last time seen;
 - the AS number, AS name and country last looked up, and when;
 - total requests, and how many were forwarded and how many refused;
-- bytes in each direction;
+- body bytes of its requests, and of the responses it was sent;
 - responses by status class;
 - offences by kind: `limit`, `attack`, `rule_blocked` and `waf_blocked`. A file
   may also hold `token_refused`, which is kept, and which nothing adds to.

internal/discriminator/abuseipdb.go

@@ -0,0 +1,55 @@
+package discriminator
+
+// The heuristic abuseipdb. It reads SWW_ABUSEIPDB_KEY,
+// SWW_ABUSEIPDB_MIN_SCORE and SWW_REPUTATION_ACTION.
+
+import (
+	"context"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/bans"
+	"sneak.berlin/go/smallwebwaf/internal/ratelimit"
+	"sneak.berlin/go/smallwebwaf/internal/reputation"
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// abuseIPDB notes AbuseIPDB, as noteHit does, with the score, when its score
+// of the client is a hit, and refuses the request when SWW_REPUTATION_ACTION
+// is deny, as dnsbl does for a zone. While SWW_ABUSEIPDB_KEY is unset it does
+// nothing. A client without a score is checked in the background, by the
+// request's address, if its history counts an offence, and the request does
+// not wait for the answer. The score is then used for each address of the
+// client. Its figures are the score and SWW_ABUSEIPDB_MIN_SCORE.
+type abuseIPDB struct{}
+
+func (abuseIPDB) name() string { return "abuseipdb" }
+
+func (abuseIPDB) check(ctx context.Context, rq *request, _ time.Time) *verdict {
+	if rq.h.config.AbuseIPDBKey == "" {
+		return nil
+	}
+
+	client := rq.h.clientGroup(rq.client)
+	held, _ := rq.h.limiter.Client(client)
+	offender := held.History.Offences != ratelimit.Offences{}
+
+	score, hit := rq.h.abuseIPDB.Hit(ctx, client, rq.client, offender)
+	if !hit {
+		return nil
+	}
+
+	rq.abuseIPDBHit = true
+	rq.noteHit(bans.ReputationHit{Source: reputation.AbuseIPDBSource, Score: &score},
+		"scored by AbuseIPDB at or over SWW_ABUSEIPDB_MIN_SCORE")
+
+	if rq.h.config.ReputationAction != deny {
+		return nil
+	}
+
+	return &verdict{
+		action: requestlog.ActionDenied,
+		figures: map[string]any{
+			figureScore: score, "min_score": rq.h.config.AbuseIPDBMinScore,
+		},
+	}
+}

internal/discriminator/answertime_test.go

@@ -91,6 +91,18 @@ func TestSixteenThousandCookiesAreRefusedWellWithinTheTimeLimit(t *testing.T) {
 	}
 }
 
+func TestHeaderLimitRefusalNamesItsHeuristicCountsAndLimits(t *testing.T) {
+	t.Parallel()
+
+	s := startDiscriminator(t, "", nil)
+	req := s.question(siteA, client, "/")
+	fillHeaders(req, headerFieldLimit+1)
+	do(t, req)
+
+	wantHeuristic(t, s.out.requestLine(t), "header_limits",
+		`{"fields": 101, "fields_limit": 100, "cookies": 0, "cookies_limit": 50}`)
+}
+
 func TestEachHeaderLimitRefusesARequestOverIt(t *testing.T) {
 	t.Parallel()
 

internal/discriminator/bans.go

@@ -1,7 +1,10 @@
 package discriminator
 
+// The heuristic active_ban, and the code that makes bans, which several
+// heuristics share. active_ban reads SWW_MODE; the bans are the ban ledger's.
+
 import (
-	"net/http"
+	"context"
 	"net/netip"
 	"time"
 
@@ -18,11 +21,15 @@ func (rq *request) banResponse(action string) *refusal {
 	return &refusal{status: rq.h.config.BanResponse, action: action}
 }
 
-// banned reports whether a ban on a netblock the client is in covers the
+// activeBan refuses a client a ban covers: a ban on a netblock the client is
-// request at now, and notes for the log line when that ban ends. A
+// in, in force at now. It notes for the log line when that ban ends. A request
-// request that makes the ban permanent, or in observe mode would have,
+// that makes the ban permanent, or in observe mode would have, raises the alert
-// raises the alert for it.
+// for it. Its figures are the ban's netblock and when it ends.
-func (rq *request) banned(now time.Time) bool {
+type activeBan struct{}
+
+func (activeBan) name() string { return "active_ban" }
+
+func (activeBan) check(_ context.Context, rq *request, now time.Time) *verdict {
 	check := rq.h.ledger.Check
 	if rq.h.config.Observe {
 		check = rq.h.ledger.Find // the ban refuses nothing, and stays as it is
@@ -38,85 +45,24 @@ func (rq *request) banned(now time.Time) bool {
 		rq.alertBan(ban)
 	}
 
-	return banned
+	if !banned {
+		return nil
 	}
 
-// limitBroken counts the request for the rate limits at now, notes the
+	return &verdict{
-// client's counts for the log line, and reports whether the request takes
+		action: requestlog.ActionBanned,
-// the client over a rate limit, as its limit percentage lowers it, which
+		figures: map[string]any{
-// breaks it.
+			figureNetblock: ban.Netblock.String(), "ban_expires": rq.line.BanExpires,
-func (rq *request) limitBroken(now time.Time) bool {
+		},
-	counts, hit, over := rq.h.limiter.Count(rq.h.clientGroup(rq.client), now,
-		rq.limitPercent.percent)
-	rq.line.Counts = counts
-
-	if over {
-		rq.banForLimit(now, hit, rq.h.config.BanResponse)
-	}
-
-	return over
-}
-
-// countBytes counts the request's bytes, as countedBytes gives them, for
-// the byte limits, once its report has come, and notes the client's byte totals
-// for the log line; its requests stay there as the rate limits counted them.
-// Only an allowed request has them counted, and only one the rate limits
-// counted; in observe mode, not one that enforce mode would have refused. Bytes
-// that take the client over a byte limit, as its limit percentage for the byte
-// limits lowers it, break it; the response was passed on whole.
-func (rq *request) countBytes() {
-	if !rq.counted || rq.line.WouldAction != "" {
-		return
-	}
-
-	now := rq.h.now()
-
-	counts, hit, over := rq.h.limiter.CountBytes(rq.h.clientGroup(rq.client), now,
-		rq.countedBytes(), rq.bytesPercent.percent)
-	rq.line.Counts.MinuteBytes = counts.MinuteBytes
-	rq.line.Counts.HourBytes = counts.HourBytes
-	rq.line.Counts.DayBytes = counts.DayBytes
-
-	if over {
-		rq.banForLimit(now, hit, rq.out.status)
 	}
 }
 
-// countRefusal counts the request for the error burst once it has been
+// limitFigures are the figures of a heuristic that broke hit, a limit: its
-// answered, if discriminator refused it after a rule file match, a trap path or a
+// window, the count and the limit.
-// Core Rule Set match, and in observe mode if enforce mode would have: more than
+func limitFigures(hit ratelimit.Hit) map[string]any {
-// SWW_ERROR_BURST_THRESHOLD such refusals of the client within a minute
+	return map[string]any{
-// break a limit. A client in SWW_ALLOW_NETS, which the checks skip, is
+		figureWindow: hit.Window, figureCount: hit.Count, figureLimit: hit.Limit,
-// not counted, and nothing is while the threshold is off.
-func (rq *request) countRefusal() {
-	cfg := rq.h.config
-	if cfg.ErrorBurstThreshold == 0 {
-		return
 	}
-
-	if !rq.attack && !rq.ruleBlocked && !rq.wafBlocked {
-		return
-	}
-
-	now := rq.h.now()
-
-	hit, over := rq.h.limiter.CountRefusal(rq.h.clientGroup(rq.client), now,
-		cfg.ErrorBurstThreshold)
-	if !over {
-		return
-	}
-
-	// What the client was sent, or in observe mode would have been.
-	status := rq.out.status
-
-	switch rq.line.WouldAction {
-	case requestlog.ActionRuleBlocked, requestlog.ActionWAFBlocked:
-		status = http.StatusForbidden
-	case requestlog.ActionBanned:
-		status = cfg.BanResponse
-	}
-
-	rq.banForLimit(now, hit, status)
 }
 
 // countedBytes returns the request's bytes, once it has ended, as the

internal/discriminator/bans_test.go

@@ -228,6 +228,22 @@ func TestBannedClientIsRefusedBeforeItsCountryIsLookedUp(t *testing.T) {
 	}
 }
 
+func TestRefusalUnderABanNamesItsHeuristicNetblockAndEnd(t *testing.T) {
+	t.Parallel()
+
+	s, clk, _ := startWithClock(t, "", map[string]string{
+		rateLimitPerMinute: "1", banScopeV4Prefix: "24",
+	})
+	expires := requestlog.FormatTime(clk.Now().Add(time.Hour))
+
+	s.get(client, http.StatusOK, requestlog.ActionForward)
+	s.get(client, http.StatusForbidden, requestlog.ActionRateLimited)
+
+	line := s.get(otherClient, http.StatusForbidden, requestlog.ActionBanned)
+	wantHeuristic(t, line, "active_ban",
+		`{"netblock": "`+clientsNet+`", "ban_expires": "`+expires+`"}`)
+}
+
 func TestBanResponseAnswersEveryRefusalButTheSizeLimits(t *testing.T) {
 	t.Parallel()
 

internal/discriminator/blocklists.go

@@ -0,0 +1,41 @@
+package discriminator
+
+// The heuristic blocklists. It reads SWW_BLOCKLIST_ACTION, and the blocklists
+// through the reputation lists.
+
+import (
+	"context"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// blocklists notes the blocklists that list the client, as noteListed does,
+// and refuses the request when SWW_BLOCKLIST_ACTION is deny. Being limit, it
+// lowers the client's limits instead (see limitPercentages), and being log, it
+// does nothing more. Its figure is the blocklists that list the client.
+type blocklists struct{}
+
+func (blocklists) name() string { return "blocklists" }
+
+func (blocklists) check(_ context.Context, rq *request, _ time.Time) *verdict {
+	listedBy := rq.h.lists.ListedBy(rq.client)
+	rq.blocklisted = len(listedBy) > 0
+	rq.noteListed(listedBy, "listed by a blocklist")
+
+	if !rq.refusedByBlocklists(listedBy) {
+		return nil
+	}
+
+	return &verdict{
+		action:  requestlog.ActionDenied,
+		figures: map[string]any{"lists": listedBy},
+	}
+}
+
+// refusedByBlocklists reports whether the heuristic blocklists refuses a
+// client, given listedBy, the blocklists that list it: it does when one lists
+// it and SWW_BLOCKLIST_ACTION is deny.
+func (rq *request) refusedByBlocklists(listedBy []string) bool {
+	return len(listedBy) > 0 && rq.h.config.BlocklistAction == deny
+}

internal/discriminator/bytelimits.go

@@ -0,0 +1,43 @@
+package discriminator
+
+// The heuristic byte_limits. It reads SWW_BYTES_COUNT, and the byte limits
+// and biased thresholds through the limiter and limitPercentages.
+
+import (
+	"context"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// byteLimits counts the request's bytes, as countedBytes gives them, for the
+// byte limits, once its report has come, and notes the client's byte totals for
+// the log line; its requests stay there as the rate limits counted them. Only
+// an allowed request has them counted, and only one the rate limits counted; in
+// observe mode, not one that enforce mode would have refused. Bytes that take
+// the client over a byte limit, as its limit percentage for the byte limits
+// lowers it, break it, which bans the client; the response was passed on whole.
+// Its figures are the window, the count and the limit.
+type byteLimits struct{}
+
+func (byteLimits) name() string { return "byte_limits" }
+
+func (byteLimits) check(_ context.Context, rq *request, now time.Time) *verdict {
+	if !rq.counted || rq.line.WouldAction != "" {
+		return nil
+	}
+
+	counts, hit, over := rq.h.limiter.CountBytes(rq.h.clientGroup(rq.client), now,
+		rq.countedBytes(), rq.bytesPercent.percent)
+	rq.line.Counts.MinuteBytes = counts.MinuteBytes
+	rq.line.Counts.HourBytes = counts.HourBytes
+	rq.line.Counts.DayBytes = counts.DayBytes
+
+	if !over {
+		return nil
+	}
+
+	rq.banForLimit(now, hit, rq.out.status)
+
+	return &verdict{action: requestlog.ActionBanned, figures: limitFigures(hit)}
+}

internal/discriminator/bytelimits_test.go

@@ -98,6 +98,21 @@ func TestEachByteLimitBansOnceTheResponseHasEnded(t *testing.T) {
 	}
 }
 
+func TestByteLimitBanNamesItsHeuristicWindowCountAndLimit(t *testing.T) {
+	t.Parallel()
+
+	s, _ := startWithAnswers(t, map[string]string{bytesLimitPerMinute: byteLimit})
+
+	line, _ := s.download()
+	wantAfterAnswerHeuristic(t, line, "", "null")
+
+	// Two answers of 70 bytes are over the limit of 99.
+	line, _ = s.download()
+	wantHeuristic(t, line, "", "null")
+	wantAfterAnswerHeuristic(t, line, "byte_limits",
+		`{"window": "minute", "count": 140, "limit": 99}`)
+}
+
 func TestResponseOverAByteLimitByItselfIsPassedOnWhole(t *testing.T) {
 	t.Parallel()
 

internal/discriminator/coreruleset.go

@@ -1,6 +1,10 @@
 package discriminator
 
+// The heuristic core_rule_set. It reads SWW_WAF_MODE, SWW_WAF_EXEMPT_PATHS
+// and SWW_WAF_ANOMALY_THRESHOLD.
+
 import (
+	"context"
 	"time"
 
 	"sneak.berlin/go/smallwebwaf/internal/alerts"
@@ -9,18 +13,21 @@ import (
 	"sneak.berlin/go/smallwebwaf/internal/waf"
 )
 
-// checkCoreRuleSet inspects the request with the Core Rule Set, unless
+// coreRuleSet inspects the request with the Core Rule Set, unless
 // SWW_WAF_MODE is off or SWW_WAF_EXEMPT_PATHS exempts its path, as
 // pathExempt decides, and notes the rules it matched and its score in the
 // log line, and the rules in the metrics. A score at or over
 // SWW_WAF_ANOMALY_THRESHOLD is a match: it raises the waf_block alert,
-// and in block mode refuses the request, which is an offence its client's
+// and in block mode refuses the request with 403, which is an offence its
-// history counts, and so returns ActionWAFBlocked. It returns "" for a
+// client's history counts. Its figures are the score and the threshold.
-// request it does not refuse.
+type coreRuleSet struct{}
-func (rq *request) checkCoreRuleSet() string {
+
+func (coreRuleSet) name() string { return "core_rule_set" }
+
+func (coreRuleSet) check(_ context.Context, rq *request, _ time.Time) *verdict {
 	cfg := rq.h.config
 	if cfg.WAFMode == config.WAFModeOff || pathExempt(rq.in, cfg.WAFExemptPaths) {
-		return ""
+		return nil
 	}
 
 	start := time.Now()
@@ -40,18 +47,21 @@ func (rq *request) checkCoreRuleSet() string {
 
 	threshold := cfg.WAFAnomalyThreshold
 	if threshold == 0 || result.Score < threshold {
-		return ""
+		return nil
 	}
 
 	rq.alertWAFBlock(result)
 
 	if cfg.WAFMode == config.WAFModeDetect {
-		return ""
+		return nil
 	}
 
 	rq.wafBlocked = true
 
-	return requestlog.ActionWAFBlocked
+	return &verdict{
+		action:  requestlog.ActionWAFBlocked,
+		figures: map[string]any{figureScore: result.Score, "threshold": threshold},
+	}
 }
 
 // alertWAFBlock raises the waf_block alert for the request, which the Core

internal/discriminator/coreruleset_test.go

@@ -110,6 +110,16 @@ func TestCoreRuleSetRefusesAttacksInBlockModeAndOnlyLogsThemInDetectMode(t *test
 	}
 }
 
+func TestCoreRuleSetRefusalNamesItsHeuristicScoreAndThreshold(t *testing.T) {
+	t.Parallel()
+
+	s, _, _ := startWithClock(t, "", map[string]string{wafMode: block})
+
+	line := s.request(client, sqlInjection, http.StatusForbidden,
+		requestlog.ActionWAFBlocked)
+	wantHeuristic(t, line, "core_rule_set", `{"score": 5, "threshold": 5}`)
+}
+
 func TestOrdinaryRequestIsInspectedAndPassed(t *testing.T) {
 	t.Parallel()
 

internal/discriminator/countries.go

@@ -1,21 +1,42 @@
 package discriminator
 
+// The heuristic countries. It reads SWW_DENIED_COUNTRIES and
+// SWW_EXCLUSIVELY_ALLOWED_COUNTRIES.
+
 import (
+	"context"
 	"slices"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
 )
 
-// countryDenied reports whether the country lists refuse the request, by
+// countries refuses a request by the client's country as it was looked up. A
-// the client's country as it was looked up. A client without a country,
+// client without a country, or whose country cannot be found, is refused only
-// or whose country cannot be found, is refused only by
+// by SWW_EXCLUSIVELY_ALLOWED_COUNTRIES. Its figures are the country and the
-// SWW_EXCLUSIVELY_ALLOWED_COUNTRIES.
+// setting that refused it.
-func (rq *request) countryDenied() bool {
+type countries struct{}
+
+func (countries) name() string { return "countries" }
+
+func (countries) check(_ context.Context, rq *request, _ time.Time) *verdict {
 	denied := rq.h.config.DeniedCountries
 	allowed := rq.h.config.ExclusivelyAllowedCountries
 	country := rq.line.Country
 
-	if slices.Contains(denied, country) {
+	var setting string
-		return true
+
+	switch {
+	case slices.Contains(denied, country):
+		setting = "SWW_DENIED_COUNTRIES"
+	case len(allowed) > 0 && !slices.Contains(allowed, country):
+		setting = "SWW_EXCLUSIVELY_ALLOWED_COUNTRIES"
+	default:
+		return nil
 	}
 
-	return len(allowed) > 0 && !slices.Contains(allowed, country)
+	return &verdict{
+		action:  requestlog.ActionCountryDenied,
+		figures: map[string]any{"country": country, "setting": setting},
+	}
 }

internal/discriminator/countries_test.go

@@ -90,6 +90,23 @@ func TestCountryLists(t *testing.T) {
 	}
 }
 
+func TestCountryRefusalNamesItsHeuristicCountryAndSetting(t *testing.T) {
+	t.Parallel()
+
+	geojsURL, _ := startGeoJS(t)
+	s, _, _ := startWithClock(t, geojsURL, map[string]string{
+		lookupTimeout: "1h", deniedCountries: "kp", allowedCountries: "de",
+	})
+
+	line := s.get(fromKP, http.StatusForbidden, requestlog.ActionCountryDenied)
+	wantHeuristic(t, line, "countries",
+		`{"country": "KP", "setting": "SWW_DENIED_COUNTRIES"}`)
+
+	line = s.get(unplaced, http.StatusForbidden, requestlog.ActionCountryDenied)
+	wantHeuristic(t, line, "countries",
+		`{"country": "", "setting": "SWW_EXCLUSIVELY_ALLOWED_COUNTRIES"}`)
+}
+
 func TestCountryRefusalKeepsARequestFromTheApp(t *testing.T) {
 	t.Parallel()
 

internal/discriminator/crowdsec.go

@@ -0,0 +1,49 @@
+package discriminator
+
+// The heuristic crowdsec. It reads SWW_CROWDSEC_LAPI_URL, and the decisions
+// through the reputation lists.
+
+import (
+	"context"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/bans"
+	"sneak.berlin/go/smallwebwaf/internal/reputation"
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// crowdSec refuses a request when a decision of the CrowdSec decision list on
+// the client is in force at now. It then notes the list, as noteHit does, and
+// bans the client until that decision ends. Its figures are the decision's
+// scenario and when it ends.
+type crowdSec struct{}
+
+func (crowdSec) name() string { return "crowdsec" }
+
+func (crowdSec) check(_ context.Context, rq *request, now time.Time) *verdict {
+	decision, listed := rq.crowdSecDecisionInForce(now)
+	if !listed {
+		return nil
+	}
+
+	rq.noteHit(bans.ReputationHit{Source: rq.h.config.CrowdSecDecisionsURL},
+		"listed by the CrowdSec decision list")
+	rq.banForCrowdSec(now, decision)
+
+	return &verdict{
+		action: requestlog.ActionBanned,
+		figures: map[string]any{
+			"scenario": decision.Scenario,
+			"expires":  requestlog.FormatTime(decision.Expires),
+		},
+	}
+}
+
+// crowdSecDecisionInForce returns the decision of the CrowdSec decision list on
+// the client that is in force at now, and reports whether there is one, which
+// the heuristic crowdsec refuses the request for.
+func (rq *request) crowdSecDecisionInForce(
+	now time.Time,
+) (reputation.Decision, bool) {
+	return rq.h.lists.CrowdSecDecision(rq.client, now)
+}

internal/discriminator/crowdsec_test.go

@@ -95,6 +95,21 @@ func TestClientTheCrowdSecDecisionListListsIsBannedUntilTheDecisionEnds(t *testi
 		float64(fetched.Unix()))
 }
 
+func TestCrowdSecBanNamesItsHeuristicScenarioAndEnd(t *testing.T) {
+	t.Parallel()
+
+	s, clk, server, _ := startWithAlerts(t, map[string]string{
+		crowdSecURL: lapi, crowdSecKey: bouncerKey,
+	})
+	loadDecisions(t, server, clk.Now(), `[{"duration": "4h0m0s", `+
+		`"scenario": "crowdsecurity/ssh-bf", "scope": "Ip", "type": "ban", `+
+		`"value": "`+client+`"}]`)
+
+	line := s.get(client, http.StatusForbidden, requestlog.ActionBanned)
+	wantHeuristic(t, line, "crowdsec", `{"scenario": "crowdsecurity/ssh-bf", `+
+		`"expires": "`+requestlog.FormatTime(clk.Now().Add(4*time.Hour))+`"}`)
+}
+
 func TestEndedCrowdSecDecisionNoLongerBansThoughTheCopyStillHoldsIt(t *testing.T) {
 	t.Parallel()
 

internal/discriminator/denynets.go

@@ -0,0 +1,29 @@
+package discriminator
+
+// The heuristic deny_nets. It reads SWW_DENY_NETS.
+
+import (
+	"context"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// denyNets refuses a client in SWW_DENY_NETS. Its figure is the netblock of
+// the setting the client is in.
+type denyNets struct{}
+
+func (denyNets) name() string { return "deny_nets" }
+
+func (denyNets) check(_ context.Context, rq *request, _ time.Time) *verdict {
+	for _, netblock := range rq.h.config.DenyNets {
+		if netblock.Contains(rq.client) {
+			return &verdict{
+				action:  requestlog.ActionDenied,
+				figures: map[string]any{figureNetblock: netblock.String()},
+			}
+		}
+	}
+
+	return nil
+}

internal/discriminator/discriminator.go

@@ -2,6 +2,7 @@ package discriminator
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -52,13 +53,6 @@ const ReportPath = "/_sww/report"
 // QuestionBodyMaxBytes is the most of a visitor's body a question carries.
 const QuestionBodyMaxBytes = 64 << 10
 
-// The header limits: the most cookies and header fields a request may have
-// when the checks reach the Core Rule Set. Real clients never reach them.
-const (
-	maxCookies      = 50
-	maxHeaderFields = 100
-)
-
 const (
 	// maxAwaitingReport is the most answered requests held until their
 	// reports come. Past it, the request answered longest ago is dropped.
@@ -148,48 +142,20 @@ func (h *handler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
 
 // refusedWhateverTheAnswer reports whether the request is one that does not wait
 // for GeoJS, since the checks will refuse it at now whatever GeoJS answers about
-// its client: it is over the header limits, a blocklist lists its client while
+// its client: the heuristic crowdsec, header_limits, blocklists or rate_limits
-// SWW_BLOCKLIST_ACTION is deny, the CrowdSec decision list lists it, or it is
+// will refuse it, as a function in that heuristic's file says. In observe mode,
-// over a whole rate limit. In observe mode, which refuses nothing, none is.
+// which refuses nothing, none is. rproxy lets a request through when
-// rproxy lets a request through when discriminator answers too late.
+// discriminator answers too late.
 func (rq *request) refusedWhateverTheAnswer(now time.Time) bool {
 	if rq.h.config.Observe {
 		return false
 	}
 
-	lists := rq.h.lists
+	_, decided := rq.crowdSecDecisionInForce(now)
-	_, decided := lists.CrowdSecDecision(rq.client, now)
 
-	return decided || rq.overHeaderLimits() ||
+	return decided || overHeaderLimits(rq.headerCounts()) ||
-		(rq.h.config.BlocklistAction == deny && len(lists.ListedBy(rq.client)) > 0) ||
+		rq.refusedByBlocklists(rq.h.lists.ListedBy(rq.client)) ||
-		(rq.countedByRateLimits() &&
+		rq.overAWholeRateLimit(now)
-			rq.h.limiter.OverARateLimit(rq.h.clientGroup(rq.client), now))
-}
-
-// overHeaderLimits reports whether discriminator refuses the request for its
-// headers: it has more than maxHeaderFields fields, each value of a header sent
-// twice being one, or more than maxCookies cookies, each part of a Cookie header
-// between semicolons being one unless it is empty or only spaces and tabs. The
-// Core Rule Set runs each of its rules on each cookie, so that thousands of small
-// cookies keep it busy for over a second. It trims only spaces and control
-// characters, of which a header value can hold only tabs, so a part of Unicode
-// spaces is a cookie to it.
-func (rq *request) overHeaderLimits() bool {
-	fields, cookies := 0, 0
-
-	for _, values := range rq.in.Header {
-		fields += len(values)
-	}
-
-	for _, value := range rq.in.Header.Values("Cookie") {
-		for part := range strings.SplitSeq(value, ";") {
-			if strings.Trim(part, " \t") != "" {
-				cookies++
-			}
-		}
-	}
-
-	return fields > maxHeaderFields || cookies > maxCookies
 }
 
 // answerQuestion runs the checks on the visitor's request a question
@@ -220,18 +186,18 @@ func (h *handler) answerQuestion(w http.ResponseWriter, r *http.Request) {
 
 	if refused == nil && !underOwnPrefix {
 		// In observe mode, what enforce mode would have refused counts here.
-		defer rq.countRefusal()
+		defer rq.runAfterAnswer(r.Context())
 
 		rq.allow()
 
 		return
 	}
 
-	// The last deferred runs first: countRefusal before addToHistory, so that a
+	// The last deferred runs first: runAfterAnswer before addToHistory, so that
-	// broken error burst is in the client's history.
+	// a broken refusal limit is in the client's history.
 	defer rq.addToHistory()
 	defer rq.countAnomalies()
-	defer rq.countRefusal()
+	defer rq.runAfterAnswer(r.Context())
 
 	if refused != nil {
 		rq.deny(*refused)
@@ -413,7 +379,7 @@ func (h *handler) takeReport(w http.ResponseWriter, r *http.Request) {
 	// Remove reports true to one report only, of several that name the id. A
 	// question still being handled is held as nil.
 	if found && held != nil && h.awaitingReport.Remove(report.RequestID) {
-		h.finishWith(held, report)
+		h.finishWith(r.Context(), held, report)
 	}
 
 	w.WriteHeader(http.StatusNoContent)
@@ -476,11 +442,11 @@ func (report *Report) checkRanges() error {
 
 // finishWith finishes a request with what its report says, and writes the
 // request's last log line, which holds its decision's fields and the report's.
-// For an allowed request it first counts the request's bytes, and adds it to its
+// For an allowed request it first runs the heuristics of onReport, and adds the
-// client's history, the anomaly counters and the metrics, with the status and
+// request to its client's history, the anomaly counters and the metrics, with
-// bytes the report gives. A denied request's counting was finished at its
+// the status and bytes the report gives. A denied request's counting was
-// decision.
+// finished at its decision. ctx is the report's own context.
-func (h *handler) finishWith(held *decision, report Report) {
+func (h *handler) finishWith(ctx context.Context, held *decision, report Report) {
 	rq := &request{
 		decision:     *held,
 		h:            h,
@@ -497,7 +463,7 @@ func (h *handler) finishWith(held *decision, report Report) {
 	if rq.allowed {
 		rq.line.DurationUpstreamTotal = &report.DurationUpstreamTotal
 
-		rq.countBytes()
+		rq.runBanningHeuristics(ctx, onReport())
 		rq.countAnomalies()
 		rq.addToHistory()
 

internal/discriminator/dnsbl.go

@@ -0,0 +1,36 @@
+package discriminator
+
+// The heuristic dnsbl. It reads SWW_REPUTATION_ACTION, and the DNSBL zones'
+// verdicts through the DNSBL checker.
+
+import (
+	"context"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// dnsbl notes the DNSBL zones whose verdict lists the client, as noteListed
+// does, and refuses the request when SWW_REPUTATION_ACTION is deny. Being
+// limit, it lowers the client's limits instead (see limitPercentages), and
+// being log, it does nothing more. A zone without a verdict on the client is
+// asked about it in the background, and the request does not wait for the
+// answer. Its figure is the zones that list the client, their keys masked.
+type dnsbl struct{}
+
+func (dnsbl) name() string { return "dnsbl" }
+
+func (dnsbl) check(ctx context.Context, rq *request, _ time.Time) *verdict {
+	listedBy := rq.h.dnsbl.ListedBy(ctx, rq.client)
+	rq.dnsblListed = len(listedBy) > 0
+	rq.noteListed(listedBy, "listed by a DNSBL zone")
+
+	if !rq.dnsblListed || rq.h.config.ReputationAction != deny {
+		return nil
+	}
+
+	return &verdict{
+		action:  requestlog.ActionDenied,
+		figures: map[string]any{"zones": listedBy},
+	}
+}

internal/discriminator/headerlimits.go

@@ -0,0 +1,71 @@
+package discriminator
+
+// The heuristic header_limits. It reads no setting: its limits are fixed.
+
+import (
+	"context"
+	"strings"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// The header limits: the most cookies and header fields a request may have
+// when the checks reach the Core Rule Set. Real clients never reach them.
+const (
+	maxCookies      = 50
+	maxHeaderFields = 100
+)
+
+// headerLimits refuses a request over the header limits, with 431. Its figures
+// are the request's header fields and cookies, and the limit of each.
+type headerLimits struct{}
+
+func (headerLimits) name() string { return "header_limits" }
+
+func (headerLimits) check(_ context.Context, rq *request, _ time.Time) *verdict {
+	fields, cookies := rq.headerCounts()
+	if !overHeaderLimits(fields, cookies) {
+		return nil
+	}
+
+	return &verdict{
+		action: requestlog.ActionTooLarge,
+		figures: map[string]any{
+			"fields": fields, "fields_limit": maxHeaderFields,
+			"cookies": cookies, "cookies_limit": maxCookies,
+		},
+	}
+}
+
+// overHeaderLimits reports whether a request with these counts of header fields
+// and cookies, as headerCounts counts them, is over the header limits: it has
+// more than maxHeaderFields fields or more than maxCookies cookies.
+func overHeaderLimits(fields, cookies int) bool {
+	return fields > maxHeaderFields || cookies > maxCookies
+}
+
+// headerCounts returns the request's header fields, each value of a header sent
+// twice being one, and its cookies, each part of a Cookie header between
+// semicolons being one unless it is empty or only spaces and tabs. The Core
+// Rule Set runs each of its rules on each cookie, so that thousands of small
+// cookies keep it busy for over a second. It trims only spaces and control
+// characters, of which a header value can hold only tabs, so a part of Unicode
+// spaces is a cookie to it.
+func (rq *request) headerCounts() (int, int) {
+	fields, cookies := 0, 0
+
+	for _, values := range rq.in.Header {
+		fields += len(values)
+	}
+
+	for _, value := range rq.in.Header.Values("Cookie") {
+		for part := range strings.SplitSeq(value, ";") {
+			if strings.Trim(part, " \t") != "" {
+				cookies++
+			}
+		}
+	}
+
+	return fields, cookies
+}

internal/discriminator/heuristics.go

@@ -0,0 +1,134 @@
+package discriminator
+
+import (
+	"context"
+	"time"
+)
+
+// To add a heuristic, write one file with a type that implements heuristic,
+// and add one line for it to one of the lists below. Its settings go in
+// config.Config, and the comment at the top of its file names them.
+
+// heuristic is one of the checks that decide what discriminator does with a
+// request.
+type heuristic interface {
+	// name is the heuristic's name, which the log line gives.
+	name() string
+	// check looks at the request at now, and returns nil to let it through.
+	// ctx is the request's own context.
+	check(ctx context.Context, rq *request, now time.Time) *verdict
+}
+
+// verdict is a heuristic refusing a request, or banning its client: the action
+// the log line names, one of requestlog's Action constants, and the figures the
+// heuristic acted on, by name, such as the count and the limit.
+type verdict struct {
+	action  string
+	figures map[string]any
+}
+
+// The names of the figures that more than one heuristic gives.
+const (
+	figureWindow   = "window"
+	figureCount    = "count"
+	figureLimit    = "limit"
+	figureScore    = "score"
+	figureNetblock = "netblock"
+)
+
+// The lists of heuristics, each in the order its heuristics run. A client in
+// SWW_ALLOW_NETS skips them all.
+
+// beforeLookup are the heuristics that run before the client's AS number and
+// country are looked up, so that a client they refuse is not looked up.
+func beforeLookup() []heuristic {
+	return []heuristic{
+		denyNets{},
+		activeBan{},
+	}
+}
+
+// afterLookup are the heuristics that run once the lookup has been made.
+func afterLookup() []heuristic {
+	return []heuristic{
+		countries{},
+		blocklists{},
+		crowdSec{},
+		dnsbl{},
+		abuseIPDB{},
+		rateLimits{},
+		trapPaths{},
+		ruleFiles{},
+		headerLimits{},
+		coreRuleSet{},
+	}
+}
+
+// afterAnswer are the heuristics that run on a request once discriminator has
+// answered it, before its decision line is written. They ban the client, and do
+// not change what the request got: their verdict's action is not used.
+func afterAnswer() []heuristic {
+	return []heuristic{
+		refusalLimit{},
+	}
+}
+
+// onReport are the heuristics that run on an allowed request when rproxy's
+// report of it comes, before its report line is written. Like afterAnswer, they
+// only ban.
+func onReport() []heuristic {
+	return []heuristic{
+		byteLimits{},
+	}
+}
+
+// runAfterAnswer runs the heuristics of afterAnswer on the request, at the time
+// it is called.
+func (rq *request) runAfterAnswer(ctx context.Context) {
+	rq.runBanningHeuristics(ctx, afterAnswer())
+}
+
+// runHeuristics runs list, beforeLookup or afterLookup, on the request at now,
+// and returns the first verdict, or nil when none gives one. The log line's
+// heuristic and heuristic_figures name the heuristic that gave it, and give its
+// figures. ctx is the request's own context.
+func (rq *request) runHeuristics(
+	ctx context.Context, list []heuristic, now time.Time,
+) *verdict {
+	name, v := rq.firstVerdict(ctx, list, now)
+	if v != nil {
+		rq.line.Heuristic = name
+		rq.line.HeuristicFigures = v.figures
+	}
+
+	return v
+}
+
+// runBanningHeuristics runs list, afterAnswer or onReport, on the request at
+// the time it is called. The log line's after_answer_heuristic and
+// after_answer_heuristic_figures name the heuristic that gave a verdict, and
+// give its figures, so that heuristic and heuristic_figures still name the one
+// that refused the request. ctx is the request's own context.
+func (rq *request) runBanningHeuristics(ctx context.Context, list []heuristic) {
+	name, v := rq.firstVerdict(ctx, list, rq.h.now())
+	if v != nil {
+		rq.line.AfterAnswerHeuristic = name
+		rq.line.AfterAnswerHeuristicFigures = v.figures
+	}
+}
+
+// firstVerdict runs list on the request at now, in order, until a heuristic
+// gives a verdict, and returns that heuristic's name and its verdict, or "" and
+// nil when none gives one. ctx is the request's own context.
+func (rq *request) firstVerdict(
+	ctx context.Context, list []heuristic, now time.Time,
+) (string, *verdict) {
+	for _, each := range list {
+		v := each.check(ctx, rq, now)
+		if v != nil {
+			return each.name(), v
+		}
+	}
+
+	return "", nil
+}

internal/discriminator/observe_test.go

@@ -14,6 +14,18 @@ import (
 // observe is the value of SWW_MODE for observe mode.
 const observe = "observe"
 
+func TestObserveModeNamesTheHeuristicThatWouldHaveRefused(t *testing.T) {
+	t.Parallel()
+
+	s, _, _ := startWithClock(t, "", map[string]string{
+		mode: observe, denyNets: client,
+	})
+
+	line := s.get(client, http.StatusOK, requestlog.ActionForward)
+	wantWouldAction(t, line, requestlog.ActionDenied)
+	wantHeuristic(t, line, "deny_nets", `{"netblock": "`+client+`/32"}`)
+}
+
 func TestObserveModeForwardsWhatEnforceModeRefuses(t *testing.T) {
 	t.Parallel()
 

internal/discriminator/ratelimits.go

@@ -0,0 +1,65 @@
+package discriminator
+
+// The heuristic rate_limits. It reads SWW_RATE_LIMIT_EXEMPT_NETS,
+// SWW_RATE_LIMIT_EXEMPT_PATHS and SWW_BAN_RESPONSE, and the rate limits and
+// biased thresholds through the limiter and limitPercentages.
+
+import (
+	"context"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// rateLimits counts the request for the rate limits, unless
+// countedByRateLimits exempts it, by the client's limit percentages, which it
+// notes for the log line with the client's counts. A request that takes the
+// client over a rate limit, as its limit percentage lowers it, breaks it: it is
+// refused, and bans the client. A request exempt from the rate limits is exempt
+// from the byte limits too. Its figures are the window, the count and the
+// limit.
+type rateLimits struct{}
+
+func (rateLimits) name() string { return "rate_limits" }
+
+func (rateLimits) check(_ context.Context, rq *request, now time.Time) *verdict {
+	rq.counted = rq.countedByRateLimits()
+	if !rq.counted {
+		return nil
+	}
+
+	rq.limitPercent, rq.bytesPercent = rq.limitPercentages()
+	rq.line.LimitPercent, rq.line.LimitPercentSetting = rq.limitPercent.logged()
+	rq.line.BytesPercent, rq.line.BytesPercentSetting = rq.bytesPercent.logged()
+
+	counts, hit, over := rq.h.limiter.Count(rq.h.clientGroup(rq.client), now,
+		rq.limitPercent.percent)
+	rq.line.Counts = counts
+
+	if !over {
+		return nil
+	}
+
+	rq.banForLimit(now, hit, rq.h.config.BanResponse)
+
+	return &verdict{action: requestlog.ActionRateLimited, figures: limitFigures(hit)}
+}
+
+// overAWholeRateLimit reports whether the heuristic rate_limits will refuse the
+// request at now, whatever its client's limit percentages: the rate limits
+// count it, and one more request from its client takes it over a whole rate
+// limit. It counts nothing.
+func (rq *request) overAWholeRateLimit(now time.Time) bool {
+	return rq.countedByRateLimits() &&
+		rq.h.limiter.OverARateLimit(rq.h.clientGroup(rq.client), now)
+}
+
+// countedByRateLimits reports whether the rate limits count the request: its
+// client is not in SWW_RATE_LIMIT_EXEMPT_NETS, and its path is not exempt
+// under SWW_RATE_LIMIT_EXEMPT_PATHS.
+func (rq *request) countedByRateLimits() bool {
+	cfg := rq.h.config
+
+	return !isInside(rq.client, cfg.RateLimitExemptNets) &&
+		!pathExempt(rq.in, cfg.RateLimitExemptPaths)
+}

internal/discriminator/ratelimits_test.go

@@ -72,6 +72,17 @@ func TestRateLimitRefusesBeforeTheApp(t *testing.T) {
 	}
 }
 
+func TestRateLimitRefusalNamesItsHeuristicWindowCountAndLimit(t *testing.T) {
+	t.Parallel()
+
+	s, _, _ := startWithClock(t, "", map[string]string{rateLimitPerMinute: "1"})
+
+	s.get(client, http.StatusOK, requestlog.ActionForward)
+
+	line := s.get(client, http.StatusForbidden, requestlog.ActionRateLimited)
+	wantHeuristic(t, line, "rate_limits", `{"window": "minute", "count": 2, "limit": 1}`)
+}
+
 func TestIPv6GroupPrefixSetsTheClientTheLimitsCount(t *testing.T) {
 	t.Parallel()
 

internal/discriminator/refusallimit.go

@@ -0,0 +1,54 @@
+package discriminator
+
+// The heuristic refusal_limit. It reads SWW_ERROR_BURST_THRESHOLD and
+// SWW_BAN_RESPONSE.
+
+import (
+	"context"
+	"net/http"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// refusalLimit counts the request once it has been answered, if discriminator
+// refused it after a rule file match, a trap path or a Core Rule Set match, and
+// in observe mode if enforce mode would have: more than
+// SWW_ERROR_BURST_THRESHOLD such refusals of the client within a minute break
+// a limit, which bans the client. A client in SWW_ALLOW_NETS, which the
+// heuristics skip, is not counted, and nothing is while the threshold is off.
+// Its figures are the window, the count and the limit.
+type refusalLimit struct{}
+
+func (refusalLimit) name() string { return "refusal_limit" }
+
+func (refusalLimit) check(_ context.Context, rq *request, now time.Time) *verdict {
+	cfg := rq.h.config
+	if cfg.ErrorBurstThreshold == 0 {
+		return nil
+	}
+
+	if !rq.attack && !rq.ruleBlocked && !rq.wafBlocked {
+		return nil
+	}
+
+	hit, over := rq.h.limiter.CountRefusal(rq.h.clientGroup(rq.client), now,
+		cfg.ErrorBurstThreshold)
+	if !over {
+		return nil
+	}
+
+	// What the client was sent, or in observe mode would have been.
+	status := rq.out.status
+
+	switch rq.line.WouldAction {
+	case requestlog.ActionRuleBlocked, requestlog.ActionWAFBlocked:
+		status = http.StatusForbidden
+	case requestlog.ActionBanned:
+		status = cfg.BanResponse
+	}
+
+	rq.banForLimit(now, hit, status)
+
+	return &verdict{action: requestlog.ActionBanned, figures: limitFigures(hit)}
+}

internal/discriminator/refusallimit_test.go

@@ -106,6 +106,25 @@ func TestErrorBurstBreaksAtOneOverTheThreshold(t *testing.T) {
 	}
 }
 
+func TestRefusalLimitBanNamesItsHeuristicWindowCountAndLimit(t *testing.T) {
+	t.Parallel()
+
+	s, _, _ := startForErrorBurst(t, "1", nil)
+
+	const ruleFigures = `{"rule_id": "blocked", "rule_action": "block"}`
+
+	line := blockRule.send(s, client)
+	wantHeuristic(t, line, "rule_files", ruleFigures)
+	wantAfterAnswerHeuristic(t, line, "", "null")
+
+	// The second refusal is the request refusal_limit bans for, and the line
+	// names both.
+	line = blockRule.send(s, client)
+	wantHeuristic(t, line, "rule_files", ruleFigures)
+	wantAfterAnswerHeuristic(t, line, "refusal_limit",
+		`{"window": "minute", "count": 2, "limit": 1}`)
+}
+
 func TestErrorBurstBanNotesHistoryAndMetrics(t *testing.T) {
 	t.Parallel()
 

internal/discriminator/reputation.go

@@ -1,91 +1,17 @@
 package discriminator
 
-import (
+// What the reputation heuristics share: blocklists, crowdsec, dnsbl and
-	"context"
+// abuseipdb.
-	"time"
 
+import (
 	"sneak.berlin/go/smallwebwaf/internal/alerts"
 	"sneak.berlin/go/smallwebwaf/internal/bans"
-	"sneak.berlin/go/smallwebwaf/internal/ratelimit"
-	"sneak.berlin/go/smallwebwaf/internal/reputation"
 )
 
 // deny is the SWW_BLOCKLIST_ACTION and the SWW_REPUTATION_ACTION that
 // refuses the requests of a client a source lists.
 const deny = "deny"
 
-// blocklistDenied notes the blocklists that list the client, as
-// noteListed does, and reports whether SWW_BLOCKLIST_ACTION, being deny,
-// refuses the request. Being limit, it lowers the client's limits instead
-// (see limitPercentages), and being log, it does nothing more.
-func (rq *request) blocklistDenied() bool {
-	listedBy := rq.h.lists.ListedBy(rq.client)
-	rq.blocklisted = len(listedBy) > 0
-	rq.noteListed(listedBy, "listed by a blocklist")
-
-	return rq.blocklisted && rq.h.config.BlocklistAction == deny
-}
-
-// crowdSecBanned reports whether a decision of the CrowdSec decision list
-// on the client is in force at now. If one is, it notes the list, as
-// noteHit does, and bans the client until that decision ends.
-func (rq *request) crowdSecBanned(now time.Time) bool {
-	decision, listed := rq.h.lists.CrowdSecDecision(rq.client, now)
-	if !listed {
-		return false
-	}
-
-	rq.noteHit(bans.ReputationHit{Source: rq.h.config.CrowdSecDecisionsURL},
-		"listed by the CrowdSec decision list")
-	rq.banForCrowdSec(now, decision)
-
-	return true
-}
-
-// dnsblDenied notes the DNSBL zones whose verdict lists the client, as
-// noteListed does, and reports whether SWW_REPUTATION_ACTION, being
-// deny, refuses the request. Being limit, it lowers the client's limits
-// instead (see limitPercentages), and being log, it does nothing more. A
-// zone without a verdict on the client is asked about it in the
-// background, and the request does not wait for the answer. ctx is the
-// request's own context.
-func (rq *request) dnsblDenied(ctx context.Context) bool {
-	listedBy := rq.h.dnsbl.ListedBy(ctx, rq.client)
-	rq.dnsblListed = len(listedBy) > 0
-	rq.noteListed(listedBy, "listed by a DNSBL zone")
-
-	return rq.dnsblListed && rq.h.config.ReputationAction == deny
-}
-
-// abuseIPDBDenied notes AbuseIPDB, as noteHit does, with the score, when
-// its score of the client is a hit, and reports whether
-// SWW_REPUTATION_ACTION, being deny, refuses the request, as dnsblDenied
-// does for a zone. While SWW_ABUSEIPDB_KEY is unset it does nothing. A
-// client without a score is checked in the background, by the request's
-// address, if its history counts an offence, and the request does not
-// wait for the answer. The score is then used for each address of the
-// client. ctx is the request's own context.
-func (rq *request) abuseIPDBDenied(ctx context.Context) bool {
-	if rq.h.config.AbuseIPDBKey == "" {
-		return false
-	}
-
-	client := rq.h.clientGroup(rq.client)
-	held, _ := rq.h.limiter.Client(client)
-	offender := held.History.Offences != ratelimit.Offences{}
-
-	score, hit := rq.h.abuseIPDB.Hit(ctx, client, rq.client, offender)
-	if !hit {
-		return false
-	}
-
-	rq.abuseIPDBHit = true
-	rq.noteHit(bans.ReputationHit{Source: reputation.AbuseIPDBSource, Score: &score},
-		"scored by AbuseIPDB at or over SWW_ABUSEIPDB_MIN_SCORE")
-
-	return rq.h.config.ReputationAction == deny
-}
-
 // noteListed notes each of sources, the URLs of the blocklists or the
 // DNSBL zones, their keys masked, that list the client, as noteHit does,
 // with reason.

internal/discriminator/reputation_test.go

@@ -118,6 +118,56 @@ func TestEachBlocklistActionForAListedAddressAndAListedNetblock(t *testing.T) {
 	}
 }
 
+func TestReputationRefusalNamesItsHeuristicAndWhatListedTheClient(t *testing.T) {
+	t.Parallel()
+
+	for _, tc := range []struct {
+		heuristic string
+		env       map[string]string
+		load      func(*testing.T, *discriminator.Server)
+		figures   string
+	}{
+		{
+			"blocklists",
+			map[string]string{blocklistURLs: dropURL, blocklistAction: actionDeny},
+			func(t *testing.T, server *discriminator.Server) {
+				t.Helper()
+				loadLists(t, server, map[string][]string{dropURL: {client}})
+			},
+			`{"lists": ["` + dropURL + `"]}`,
+		},
+		{
+			"dnsbl",
+			map[string]string{
+				dnsblZones: dnsblZone, dnsblResolver: noResolver,
+				reputationAction: actionDeny,
+			},
+			func(_ *testing.T, server *discriminator.Server) {
+				loadVerdicts(server, map[string][]string{client: {dnsblZone}})
+			},
+			`{"zones": ["` + dnsblZone + `"]}`,
+		},
+		{
+			"abuseipdb",
+			map[string]string{abuseIPDBKey: accountKey, reputationAction: actionDeny},
+			func(_ *testing.T, server *discriminator.Server) {
+				loadScores(server, map[string]int64{client: 80})
+			},
+			`{"score": 80, "min_score": 75}`,
+		},
+	} {
+		t.Run(tc.heuristic, func(t *testing.T) {
+			t.Parallel()
+
+			s, server, _ := startWithLookups(t, tc.env)
+			tc.load(t, server)
+
+			line := s.get(client, http.StatusForbidden, requestlog.ActionDenied)
+			wantHeuristic(t, line, tc.heuristic, tc.figures)
+		})
+	}
+}
+
 func TestBlocklistsComeAfterTheCountryListsAndSkipAllowNets(t *testing.T) {
 	t.Parallel()
 

internal/discriminator/request.go

@@ -160,16 +160,20 @@ func requestHeaders(r *http.Request, names []string) map[string]string {
 }
 
 // check is the one place where a request can be refused. It returns nil to let
-// the request through. The checks of checkClient decide: a request they refuse
+// the request through. The heuristics of checkClient decide: a request they
-// is answered with SWW_BAN_RESPONSE, 403 for a block rule or the Core Rule Set,
+// refuse is answered with SWW_BAN_RESPONSE, 403 for a block rule or the Core
-// or 431 for the header limits. In observe mode nothing is refused, and the log
+// Rule Set, or 431 for the header limits. In observe mode nothing is refused,
-// line names what enforce mode would have done. ctx is the request's own context.
+// and the log line names what enforce mode would have done. ctx is the
+// request's own context.
 func (rq *request) check(ctx context.Context) *refusal {
-	action := rq.checkClient(ctx)
+	v := rq.checkClient(ctx)
+	if v == nil {
+		return nil
+	}
+
+	action := v.action
 
 	switch {
-	case action == "":
-		return nil
 	case rq.h.config.Observe:
 		rq.line.WouldAction = action
 
@@ -183,35 +187,21 @@ func (rq *request) check(ctx context.Context) *refusal {
 	}
 }
 
-// checkClient runs the checks on the request's client, and returns the
+// checkClient runs the heuristics of beforeLookup on the request, then looks
-// action of the first that refuses the request, or "" when none does. A
+// up the client's AS number and country, then runs those of afterLookup. It
-// client in SWW_ALLOW_NETS skips them, and is not looked up. For any
+// returns the first verdict, or nil when none gives one. A client in
-// other client, SWW_DENY_NETS comes first, then a ban on its netblock,
+// SWW_ALLOW_NETS skips them all, and is not looked up. ctx is the request's
-// so that a client either refuses is not looked up, then the lookup of
+// own context.
-// its AS number and country, then the country lists, then the blocklists,
+func (rq *request) checkClient(ctx context.Context) *verdict {
-// then the CrowdSec decision list, which bans the client it lists, then
+	if isInside(rq.client, rq.h.config.AllowNets) {
-// the DNSBL zones' verdicts, and then AbuseIPDB's score; a request any of
+		return nil
-// them refuses is not counted for the rate limits. Then come the
-// rate limits, unless the client is in SWW_RATE_LIMIT_EXEMPT_NETS or the
-// request's path is exempt under SWW_RATE_LIMIT_EXEMPT_PATHS, so that
-// every other request is counted, each of them by the client's limit
-// percentages, then SWW_TRAP_PATHS, then the rule files, then the header
-// limits, and last the Core Rule Set. A request exempt from the rate limits
-// is exempt from the byte limits too. ctx is the request's own context.
-func (rq *request) checkClient(ctx context.Context) string {
-	cfg := rq.h.config
-	if isInside(rq.client, cfg.AllowNets) {
-		return ""
 	}
 
 	now := rq.h.now()
 
-	if isInside(rq.client, cfg.DenyNets) {
+	v := rq.runHeuristics(ctx, beforeLookup(), now)
-		return requestlog.ActionDenied
+	if v != nil {
-	}
+		return v
-
-	if rq.banned(now) {
-		return requestlog.ActionBanned
 	}
 
 	rq.lookUp(ctx, now)
@@ -220,59 +210,7 @@ func (rq *request) checkClient(ctx context.Context) string {
 	// again: a request counted at the time it arrived could be counted
 	// after a later one of its client's, which the rate limits would take
 	// for the clock set back.
-	now = rq.h.now()
+	return rq.runHeuristics(ctx, afterLookup(), rq.h.now())
-
-	if rq.countryDenied() {
-		return requestlog.ActionCountryDenied
-	}
-
-	if rq.blocklistDenied() {
-		return requestlog.ActionDenied
-	}
-
-	if rq.crowdSecBanned(now) {
-		return requestlog.ActionBanned
-	}
-
-	if rq.dnsblDenied(ctx) || rq.abuseIPDBDenied(ctx) {
-		return requestlog.ActionDenied
-	}
-
-	rq.counted = rq.countedByRateLimits()
-	if rq.counted {
-		rq.limitPercent, rq.bytesPercent = rq.limitPercentages()
-		rq.line.LimitPercent, rq.line.LimitPercentSetting = rq.limitPercent.logged()
-		rq.line.BytesPercent, rq.line.BytesPercentSetting = rq.bytesPercent.logged()
-	}
-
-	if rq.counted && rq.limitBroken(now) {
-		return requestlog.ActionRateLimited
-	}
-
-	if rq.trapPath(now) {
-		return requestlog.ActionBanned
-	}
-
-	action := rq.checkRules(now)
-	if action != "" {
-		return action
-	}
-
-	if rq.overHeaderLimits() {
-		return requestlog.ActionTooLarge
-	}
-
-	return rq.checkCoreRuleSet()
-}
-
-// countedByRateLimits reports whether the rate limits count the request: its
-// client is not in SWW_RATE_LIMIT_EXEMPT_NETS, and its path is not exempt
-// under SWW_RATE_LIMIT_EXEMPT_PATHS.
-func (rq *request) countedByRateLimits() bool {
-	cfg := rq.h.config
-
-	return !isInside(rq.client, cfg.RateLimitExemptNets) &&
-		!pathExempt(rq.in, cfg.RateLimitExemptPaths)
 }
 
 // pathExempt reports whether r is exempt under prefixes,

internal/discriminator/requestlog_test.go

@@ -1,6 +1,7 @@
 package discriminator_test
 
 import (
+	"encoding/json"
 	"io"
 	"math"
 	"net/http"
@@ -272,6 +273,48 @@ func wantTimings(t *testing.T, line logLine, want ...string) {
 	}
 }
 
+// wantHeuristic checks that line names heuristic as the one that refused the
+// request, with figures, a JSON object. A heuristic of "" with figures "null" is
+// a line that names none.
+func wantHeuristic(t *testing.T, line logLine, heuristic, figures string) {
+	t.Helper()
+
+	if line.Heuristic != heuristic ||
+		!reflect.DeepEqual(line.HeuristicFigures, decodeFigures(t, figures)) {
+		t.Errorf("log line has heuristic %q with heuristic_figures %v, want %q with %s",
+			line.Heuristic, line.HeuristicFigures, heuristic, figures)
+	}
+}
+
+// wantAfterAnswerHeuristic checks after_answer_heuristic and
+// after_answer_heuristic_figures as wantHeuristic checks heuristic and
+// heuristic_figures.
+func wantAfterAnswerHeuristic(t *testing.T, line logLine, heuristic, figures string) {
+	t.Helper()
+
+	if line.AfterAnswerHeuristic != heuristic ||
+		!reflect.DeepEqual(line.AfterAnswerHeuristicFigures, decodeFigures(t, figures)) {
+		t.Errorf("log line has after_answer_heuristic %q with "+
+			"after_answer_heuristic_figures %v, want %q with %s",
+			line.AfterAnswerHeuristic, line.AfterAnswerHeuristicFigures, heuristic, figures)
+	}
+}
+
+// decodeFigures decodes figures, a JSON object of a heuristic's figures, as
+// the log line's figures are decoded.
+func decodeFigures(t *testing.T, figures string) map[string]any {
+	t.Helper()
+
+	var decoded map[string]any
+
+	err := json.Unmarshal([]byte(figures), &decoded)
+	if err != nil {
+		t.Fatalf("decode %s: %v", figures, err)
+	}
+
+	return decoded
+}
+
 // microseconds is a timing in whole microseconds.
 func microseconds(milliseconds float64) int64 {
 	return int64(math.Round(milliseconds * 1000))

internal/discriminator/rulefiles.go

@@ -0,0 +1,55 @@
+package discriminator
+
+// The heuristic rule_files. It reads the rules of the rule files, which
+// SWW_RULES_DIR and SWW_RULES_ENABLED name.
+
+import (
+	"context"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/bans"
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+	"sneak.berlin/go/smallwebwaf/internal/rules"
+)
+
+// ruleFiles checks the request against the rules of the rule files at now,
+// notes the ids of those it matches in the log line, and refuses the request
+// when the last rule it matches refuses it: with ActionRuleBlocked for a block
+// rule, and ActionBanned for a ban rule. A ban rule bans the client's netblock
+// for a clear sign of attack, or in observe mode raises the alert for the ban
+// it would have made. Either rule's match is noted as an offence, for the
+// client's history. Its figures are the rule's id and action.
+type ruleFiles struct{}
+
+func (ruleFiles) name() string { return "rule_files" }
+
+func (ruleFiles) check(_ context.Context, rq *request, now time.Time) *verdict {
+	matched := rq.h.rules.Match(rq.in)
+
+	for _, rule := range matched {
+		rq.line.RuleIDs = append(rq.line.RuleIDs, rule.ID)
+		rq.h.metrics.RuleMatched(rule.ID, rule.Action)
+	}
+
+	if len(matched) == 0 {
+		return nil
+	}
+
+	// Only the last rule matched can refuse the request.
+	last := matched[len(matched)-1]
+	figures := map[string]any{"rule_id": last.ID, "rule_action": last.Action}
+
+	switch last.Action {
+	case rules.ActionBlock:
+		rq.ruleBlocked = true
+
+		return &verdict{action: requestlog.ActionRuleBlocked, figures: figures}
+	case rules.ActionBan:
+		rq.attack = true
+		rq.banForAttack(now, bans.Notes{RuleID: last.ID, Target: last.Target})
+
+		return &verdict{action: requestlog.ActionBanned, figures: figures}
+	default:
+		return nil
+	}
+}

internal/discriminator/rulefiles_test.go

@@ -93,6 +93,21 @@ func TestEachRuleAction(t *testing.T) {
 	s.get(client, http.StatusTooManyRequests, requestlog.ActionBanned)
 }
 
+func TestRuleRefusalNamesItsHeuristicAndTheRule(t *testing.T) {
+	t.Parallel()
+
+	s, _, _ := startWithClock(t, "", map[string]string{
+		rulesDir: writeRules(t, "noted path log ^/\n"+testRules),
+	})
+
+	line := s.request(client, "/blocked", http.StatusForbidden,
+		requestlog.ActionRuleBlocked)
+	wantHeuristic(t, line, "rule_files", `{"rule_id": "blocked", "rule_action": "block"}`)
+
+	line = s.request(client, "/.env", http.StatusForbidden, requestlog.ActionBanned)
+	wantHeuristic(t, line, "rule_files", `{"rule_id": "probe", "rule_action": "ban"}`)
+}
+
 func TestNextClearSignOfAttackAfterABanBansPermanently(t *testing.T) {
 	t.Parallel()
 

internal/discriminator/rules.go

@@ -1,64 +0,0 @@
-package discriminator
-
-import (
-	"slices"
-	"time"
-
-	"sneak.berlin/go/smallwebwaf/internal/bans"
-	"sneak.berlin/go/smallwebwaf/internal/requestlog"
-	"sneak.berlin/go/smallwebwaf/internal/rules"
-)
-
-// trapPath reports whether the request asks for a path in
-// SWW_TRAP_PATHS: its path as a path rule sees it, without the query,
-// as sent or once percent-decoded, is one of them. Such a request is a
-// clear sign of attack, as a ban rule's match is: it bans the client's
-// netblock, or in observe mode raises the alert for the ban it would have
-// made. The ban's notes give the path as sent.
-func (rq *request) trapPath(now time.Time) bool {
-	path := rules.Path(rq.in)
-	if !slices.Contains(rq.h.config.TrapPaths, path) &&
-		!slices.Contains(rq.h.config.TrapPaths, rules.DecodeOnce(path)) {
-		return false
-	}
-
-	rq.attack = true
-	rq.banForAttack(now, bans.Notes{TrapPath: path})
-
-	return true
-}
-
-// checkRules checks the request against the rules of the rule files at
-// now, notes the ids of those it matches in the log line, and returns the
-// action of the rule that refuses it, ActionRuleBlocked for a block rule
-// and ActionBanned for a ban rule, or "" when none does. A ban rule bans
-// the client's netblock for a clear sign of attack, or in observe mode
-// raises the alert for the ban it would have made. Either rule's match
-// is noted as an offence, for the client's history.
-func (rq *request) checkRules(now time.Time) string {
-	matched := rq.h.rules.Match(rq.in)
-
-	for _, rule := range matched {
-		rq.line.RuleIDs = append(rq.line.RuleIDs, rule.ID)
-		rq.h.metrics.RuleMatched(rule.ID, rule.Action)
-	}
-
-	if len(matched) == 0 {
-		return ""
-	}
-
-	// Only the last rule matched can refuse the request.
-	switch last := matched[len(matched)-1]; last.Action {
-	case rules.ActionBlock:
-		rq.ruleBlocked = true
-
-		return requestlog.ActionRuleBlocked
-	case rules.ActionBan:
-		rq.attack = true
-		rq.banForAttack(now, bans.Notes{RuleID: last.ID, Target: last.Target})
-
-		return requestlog.ActionBanned
-	default:
-		return ""
-	}
-}

internal/discriminator/staticlists_test.go

@@ -104,6 +104,15 @@ func TestDenyNetsRefuseBeforeTheLookup(t *testing.T) {
 	}
 }
 
+func TestDenyNetsRefusalNamesItsHeuristicAndNetblock(t *testing.T) {
+	t.Parallel()
+
+	s, _, _ := startWithClock(t, "", map[string]string{denyNets: clientsNet})
+
+	line := s.get(client, http.StatusForbidden, requestlog.ActionDenied)
+	wantHeuristic(t, line, "deny_nets", `{"netblock": "`+clientsNet+`"}`)
+}
+
 func TestRequestRefusedByDenyNetsIsNotCounted(t *testing.T) {
 	t.Parallel()
 

internal/discriminator/trappaths.go

@@ -0,0 +1,39 @@
+package discriminator
+
+// The heuristic trap_paths. It reads SWW_TRAP_PATHS.
+
+import (
+	"context"
+	"slices"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/bans"
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+	"sneak.berlin/go/smallwebwaf/internal/rules"
+)
+
+// trapPaths refuses a request for a path in SWW_TRAP_PATHS: its path as a
+// path rule sees it, without the query, as sent or once percent-decoded, is one
+// of them. Such a request is a clear sign of attack, as a ban rule's match is:
+// it bans the client's netblock, or in observe mode raises the alert for the
+// ban it would have made. The ban's notes give the path as sent, and so does
+// its figure.
+type trapPaths struct{}
+
+func (trapPaths) name() string { return "trap_paths" }
+
+func (trapPaths) check(_ context.Context, rq *request, now time.Time) *verdict {
+	path := rules.Path(rq.in)
+	if !slices.Contains(rq.h.config.TrapPaths, path) &&
+		!slices.Contains(rq.h.config.TrapPaths, rules.DecodeOnce(path)) {
+		return nil
+	}
+
+	rq.attack = true
+	rq.banForAttack(now, bans.Notes{TrapPath: path})
+
+	return &verdict{
+		action:  requestlog.ActionBanned,
+		figures: map[string]any{"path": path},
+	}
+}

internal/discriminator/trappaths_test.go

@@ -83,6 +83,16 @@ func TestTrapPathBansAsABanRuleDoes(t *testing.T) {
 	}
 }
 
+func TestTrapPathRefusalNamesItsHeuristicAndPath(t *testing.T) {
+	t.Parallel()
+
+	s, _, _ := startWithClock(t, "", map[string]string{trapPaths: trapPathList})
+
+	line := s.request(client, "/wp-login.php?redirect_to=x", http.StatusForbidden,
+		requestlog.ActionBanned)
+	wantHeuristic(t, line, "trap_paths", `{"path": "/wp-login.php"}`)
+}
+
 func TestTrapPathMatchesThePathDecodedOnce(t *testing.T) {
 	t.Parallel()
 

internal/requestlog/requestlog.go

@@ -131,6 +131,16 @@ type Line struct {
 	// ActionRuleBlocked, ActionWAFBlocked, or, in discriminator, ActionTooLarge
 	// for the header limits.
 	WouldAction string `json:"would_action,omitempty"`
+	// Heuristic is the name of the heuristic that refused the request, in
+	// observe mode too, and HeuristicFigures are the figures it acted on, by
+	// name.
+	Heuristic        string         `json:"heuristic,omitempty"`
+	HeuristicFigures map[string]any `json:"heuristic_figures,omitempty"`
+	// AfterAnswerHeuristic and AfterAnswerHeuristicFigures are the same for
+	// the heuristic that banned the client after discriminator answered the
+	// request.
+	AfterAnswerHeuristic        string         `json:"after_answer_heuristic,omitempty"`
+	AfterAnswerHeuristicFigures map[string]any `json:"after_answer_heuristic_figures,omitempty"` //nolint:lll // a struct tag cannot be split
 	// LimitPercent and LimitPercentSetting are, for a request the rate
 	// limits counted whose client a biased threshold gives a percentage of
 	// the rate limits below 100, that percentage and the setting that gave