Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Detect rule hook/v11 #12758

Draft
wants to merge 30 commits into
base: master
Choose a base branch
from
Draft

Conversation

victorjulien
Copy link
Member

Takes #12697 and overhauls some of the concepts. It removes the appfw idea, in favor of the following:

Firewall has 2 hooks: packet_filter and app_filter.

Allow a accept:hook. This immediately stops the processing of the
rules in this hook and continues processing in the next hook.

Hook order: packet_filter -> packet_td -> app_filter -> app_td.

Hooks

Each of the hooks has some unique properties:

packet_filter:

  • default policy is drop:packet
  • rules are processed in order
  • action scopes are explicit
  • drop or accept is immediate
  • accept:hook continues to packet_td

packet_td:

  • default policy is accept:hook
  • rules are ordered by IDS/IPS ordering logic
  • action scopes are implicit
  • actions are queued
  • continues to app_filter or alert/action finalize

app_filter:

  • default policy is drop:packet
  • rules are processed in order
  • action scopes are explicit
  • drop is immediate
  • accept is conditional on possible drop from packet_td
  • accept:hook continues to app_td, accept:packet or accept:flow
    continues to alert/action finalize

app_td:

  • default policy is accept:hook
  • rules are ordered by IDS/IPS ordering logic
  • action scopes are implicit
  • actions are queued
  • continues to alert/action finalize

The interplay between firewall and threat detect is still a bit more complex than I'd like, but here we're limited by trying to remain compatible to existing IDS/IPS rulesets.

FW rules concept with threat detect (ET) - hooks+unifiedTD

SV_BRANCH=OISF/suricata-verify#2327

Copy link

codecov bot commented Mar 12, 2025

Codecov Report

Attention: Patch coverage is 85.64426% with 205 lines in your changes missing coverage. Please review.

Project coverage is 80.89%. Comparing base (66e47a1) to head (4c10cee).

Additional details and impacted files
@@            Coverage Diff             @@
##           master   #12758      +/-   ##
==========================================
+ Coverage   80.87%   80.89%   +0.02%     
==========================================
  Files         936      937       +1     
  Lines      259748   260595     +847     
==========================================
+ Hits       210061   210804     +743     
- Misses      49687    49791     +104     
Flag Coverage Δ
fuzzcorpus 57.21% <56.06%> (-0.04%) ⬇️
livemode 19.55% <32.90%> (+0.10%) ⬆️
pcap 44.11% <39.97%> (+<0.01%) ⬆️
suricata-verify 63.82% <84.05%> (+0.11%) ⬆️
unittests 58.08% <49.29%> (-0.07%) ⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

🚀 New features to boost your workflow:
  • Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

@victorjulien victorjulien marked this pull request as draft March 12, 2025 11:29
@suricata-qa
Copy link

Information: QA ran without warnings.

Pipeline 25079

@suricata-qa
Copy link

Information: QA ran without warnings.

Pipeline 25085

Instead of having a per detection engine list of rule that couldn't be
prefiltered, put those into special "prefilter" engines.

For packet and frame rules this doesn't change much, it just removes
some hard coded logic from the detect engine.

For the packet non-prefilter rules in the "non-prefilter" special prefilter
engine, add additional filtering for the packet variant. It can prefilter on
alproto, dsize and dest port.

The frame non-prefilter rules are added to a single engine, that per
rule checks the alproto and the type.

For app-layer, there is an engine per progress value, per app-layer
protocol and per direction. This hooks app-layer non-prefilter rules
into the app inspect logic at the correct "progress" hook.

e.g. a rule like
        dns.query; bsize:1;

Negated MPM rules will also fall into this category:
        dns.query; content:!"abc";

Are part of a special "generic list" app engine for dns, at the
same progress hook as `dns.query`.

This all results in a lot fewer checks:

previous:

  --------------------------------------------------------------------------
  Date: 1/29/2025 -- 10:22:25. Sorted by: number of checks.
  --------------------------------------------------------------------------
   Num      Rule         Gid      Rev      Ticks        %      Checks   Matches  Max Ticks   Avg Ticks   Avg Match   Avg No Match
  -------- ------------ -------- -------- ------------ ------ -------- -------- ----------- ----------- ----------- --------------
  1        20           1        0        181919672    11.85  588808   221      60454       308.96      2691.46     308.07
  2        50           1        0        223455914    14.56  453104   418      61634       493.17      3902.59     490.02
  3        60           1        0        185990683    12.12  453104   418      60950       410.48      1795.40     409.20
  4        51           1        0        192436011    12.54  427028   6084     61223       450.64      2749.12     417.42
  5        61           1        0        180401533    11.75  427028   6084     61093       422.46      2177.04     397.10
  6        70           1        0        153899099    10.03  369836   0        61282       416.13      0.00        416.13
  7        71           1        0        123389405    8.04   369836   12833    44921       333.63      2430.23     258.27
  8        41           1        0        63889876     4.16   155824   12568    39138       410.01      1981.97     272.10
  9        40           1        0        64149724     4.18   155818   210      39792       411.70      4349.57     406.38
  10       10           1        0        70848850     4.62   65558    0        39544       1080.70     0.00        1080.70
  11       11           1        0        94743878     6.17   65558    32214    60547       1445.19     2616.14     313.92

this commit:

  --------------------------------------------------------------------------
  Date: 1/29/2025 -- 10:15:46. Sorted by: number of checks.
  --------------------------------------------------------------------------
   Num      Rule         Gid      Rev      Ticks        %      Checks   Matches  Max Ticks   Avg Ticks   Avg Match   Avg No Match
  -------- ------------ -------- -------- ------------ ------ -------- -------- ----------- ----------- ----------- --------------
  1        50           1        0        138776766    19.23  95920    418      167584      1446.80     3953.11     1435.83
  2        60           1        0        97988084     13.58  95920    418      182817      1021.56     1953.63     1017.48
  3        51           1        0        105318318    14.60  69838    6084     65649       1508.04     2873.38     1377.74
  4        61           1        0        89571260     12.41  69838    6084     164632      1282.56     2208.41     1194.20
  5        11           1        0        91132809     12.63  32779    32214    373569      2780.22     2785.58     2474.45
  6        10           1        0        66095303     9.16   32779    0        56704       2016.39     0.00        2016.39
  7        70           1        0        48107573     6.67   12928    0        42832       3721.19     0.00        3721.19
  8        71           1        0        32308792     4.48   12928    12833    39565       2499.13     2510.05     1025.09
  9        41           1        0        25546837     3.54   12886    12470    41479       1982.53     1980.84     2033.05
  10       40           1        0        26069992     3.61   12886    210      38495       2023.13     4330.05     1984.91
  11       20           1        0        639025       0.09   221      221      14750       2891.52     2891.52     0.00
To support hook based buffer names.
Per direction track progress to be able to have more fine grained
control over where the detection engines and logging hooks in.
Generic:
        <app_proto>:request_done and <app_proto>:response_done

Per protocol, it uses the registered progress (state) values. E.g.

        tls:client_hello_done

A rule ruleset could be:

        pass tls:client_hello_done any any -> any any (tls.sni; content:"www.google.com"; sid:21; alert;)
        drop tls:client_hello_done any any -> any any (sid:22;)

The pass rule is evaluated when the client hello is parsed, and if it
doesn't match the drop rule will be evaluated.

Registers each generic lists as "<alproto>:<progress state>:generic"
(e.g. "tls:client_hello_done:generic").

Ticket: OISF#7485.
For registration of app-layer inspection, no longer use the 'needs'
table from the script, but instead use the rule hook setting.

Ticket: OISF#4783.
tls.version isn't hooked to a specific state by default. Allow it to register at the rule hook.
WIP packet: track hooks that this packet triggers

hooks: set flow_start hook in packet

detect: pack prefilter engine struct

For future expansion of the fields.

detect/prefilter: put pkt mask in struct

In preparation for adding more pkt fields.

WIP detect: implement pkt:flow_start hook

SQUASH detect flow start hook
Firewall rules are like normal rule, with some key differences.

They are loaded separate, and first, from:

```yaml
firewall-rule-path: /etc/suricata/firewall/
firewall-rule-files:
  - fw.rules
```

Differences with regular "threat detection" rules:

1. these rules are evaluated before threat detection rules

2. these rules are evaluated in the order as they appear in the rule file

3. currently only rules specifying an explicit hook at supported

   a. as a consequence, no rules will be treated as (like) IP-only, PD-only or
      DE-only
For testing mixing firewall and threat detection rules.
@victorjulien victorjulien force-pushed the detect-rule-hook/v11 branch from fe99dfd to 626361f Compare March 13, 2025 15:53
@suricata-qa
Copy link

Information: QA ran without warnings.

Pipeline 25119

Add default drops for packet and app, add 'accept' action distinguishing
between a firewall `accept` and a IPS `pass`.

Firewall has 2 hooks: `packet_filter` and `app_filter`.

Allow a `accept:hook`. This immediately stops the processing of the
rules in this hook and continues processing in the next hook.

Hook order: `packet_filter` -> `packet_td` -> `app_filter` -> `app_td`.

Each of the hooks has some unique properties:

`packet_filter`:
- default policy is `drop`
- rules are process in order
- action scopes are explicit
- `drop` or `accept` is immediate
- `accept:hook` continues to `packet_td`

`packet_td`:
- default policy is `accept:hook`
- rules are ordered by IDS/IPS ordering logic
- action scopes are implicit
- actions are queued
- continues to `app_filter` or `alert/action finalize`

`app_filter`:
- default policy is `drop`
- rules are process in order
- action scopes are explicit
- `drop` is immediate
- `accept` is conditional on possible `drop` from `packet_td`
- `accept:hook` continues to `app_td`, `accept:packet` or `accept:flow`
  continues to `alert/action finalize`

`app_td`:
- default policy is `accept:hook`
- rules are ordered by IDS/IPS ordering logic
- action scopes are implicit
- actions are queued
- continues to `alert/action finalize`

Implementation:

During sigorder, split into packet_filter, app_filter and general td.

Spliting td into packet_td and app_td breaks expectations of existings
IDS/IPS rules.
Allow fw rules to work when in pass:flow mode.

When firewall mode is enabled, `pass:flow` will not skip the detection
engine anymore, but instead we'll process the firewall rules and then
apply the pass before inspecting threat detect rules.
Allow registering the progress as -1, which means it will be invoked
each time the app prefilters are called.
Special notation for allowing an app-layer rule to match on each app
update:

`alert http1:request_update ...`
`accept:hook tls:request_update ...`
@victorjulien victorjulien force-pushed the detect-rule-hook/v11 branch from 626361f to 2ec9062 Compare March 14, 2025 16:29
@victorjulien
Copy link
Member Author

Latest update adds:
accept:hook tls:request_update ... notation. This forces the rule to be evaluated with each packet updating the app-layer in the request direction. Similarly, there is tls:response_update. This exists for all protocols.

app-layer-state keyword: allow matching on the current state of the protocol:

accept:hook tls:client_hello_done $HOME_NET any -> $EXTERNAL_NET any (tls.sni; content:"www.google.com"; sid:104; alert;)
accept:hook tls:request_update $HOME_NET any -> $EXTERNAL_NET any (app-layer-state:<client_hello_done; alert; sid:105;)
accept:hook tls:request_update $HOME_NET any -> $EXTERNAL_NET any (app-layer-state:>client_hello_done; alert; sid:106;)

# accept all return states
accept:hook tls:response_update $EXTERNAL_NET any -> $HOME_NET any (alert; sid:205;)

(see also https://github.com/OISF/suricata-verify/pull/2327/files#diff-1dc1fca7ff47acd53dbaee11fceeb6bfb003e24029977f4ffba57d7a94addd40)

One observation:
In a private pcap we hit the stream.reassembly.depth in a HTTP POST body. Need to test what happens with this.

@@ -421,6 +433,11 @@ static inline void PacketAlertFinalizeProcessQueue(
SCLogDebug("sid:%u: is a pass rule, so break out of loop", s->id);
break;
}

// TODO we can also drop if alert is suppressed, right?
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

need a test for this:
ips sig that drops, alert is suppressed. Overrides a app_filter accept

@suricata-qa
Copy link

WARNING:

field baseline test %
SURI_TLPR1_stats_chk
.uptime 618 647 104.69%

Pipeline 25127

@suricata-qa
Copy link

Information: QA ran without warnings.

Pipeline 25137

@suricata-qa
Copy link

Information: QA ran without warnings.

Pipeline 25139

@suricata-qa
Copy link

Information: QA ran without warnings.

Pipeline 25142

@victorjulien
Copy link
Member Author

TODOs after a bit of live testing:

  • TLS statemachine interaction: we can stay in client_hello_done in case of a change cypher spec, a accept:hook tls:client_hello_done ... sni matches only on the first app update in this state.
  • it's hard to debug drops, need to log the app state (progress)

@victorjulien
Copy link
Member Author

Another thing: when crafting a ruleset it's getting pretty annoying to assign sids, wonder if we should allow skipping them for fw rules.

@catenacyber
Copy link
Contributor

To be closed for #12820 ?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Development

Successfully merging this pull request may close these issues.

3 participants