Fix Anthropic API key detector shape#7
Merged
santhsecurity merged 1 commit intoMay 29, 2026
Merged
Conversation
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (2)
✅ Files skipped from review due to trivial changes (1)
🚧 Files skipped from review as they are similar to previous changes (1)
📝 WalkthroughWalkthroughThis PR updates the Anthropic API key detector to match variable-length tokens instead of a fixed-length pattern. The regex pattern in ChangesAnthropic API Key Detector Pattern Expansion
🎯 1 (Trivial) | ⏱️ ~3 minutes
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@detectors/anthropic-api-key.toml`:
- Line 14: The current regex in detectors/anthropic-api-key.toml (the regex
entry matching "sk-ant-api03-[A-Za-z0-9_-]{80,120}") can match inside longer
tokens; update that regex to add non-token boundary anchors using negative
lookbehind and lookahead so the pattern only matches when not immediately
preceded or followed by URL-safe token characters (e.g., use (?<![A-Za-z0-9_-])
before the prefix and (?![A-Za-z0-9_-]) after the quantified class) to prevent
substring overmatching.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: f08aed82-55e8-4995-be5e-b243127cc7f2
📒 Files selected for processing (2)
crates/scanner/tests/contracts/anthropic-api-key.tomldetectors/anthropic-api-key.toml
|
|
||
| [[detector.patterns]] | ||
| regex = 'sk-ant-api03-[A-Za-z0-9_-]{93}-[A-Za-z0-9_-]{6}AA' | ||
| regex = 'sk-ant-api03-[A-Za-z0-9_-]{80,120}' |
There was a problem hiding this comment.
Add token boundaries to avoid substring overmatching.
Line 14 can match inside longer URL-safe blobs, which increases false positives. Add non-token boundaries around the key pattern.
Proposed fix
-regex = 'sk-ant-api03-[A-Za-z0-9_-]{80,120}'
+regex = '(?<![A-Za-z0-9_-])sk-ant-api03-[A-Za-z0-9_-]{80,120}(?![A-Za-z0-9_-])'📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| regex = 'sk-ant-api03-[A-Za-z0-9_-]{80,120}' | |
| regex = '(?<![A-Za-z0-9_-])sk-ant-api03-[A-Za-z0-9_-]{80,120}(?![A-Za-z0-9_-])' |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@detectors/anthropic-api-key.toml` at line 14, The current regex in
detectors/anthropic-api-key.toml (the regex entry matching
"sk-ant-api03-[A-Za-z0-9_-]{80,120}") can match inside longer tokens; update
that regex to add non-token boundary anchors using negative lookbehind and
lookahead so the pattern only matches when not immediately preceded or followed
by URL-safe token characters (e.g., use (?<![A-Za-z0-9_-]) before the prefix and
(?![A-Za-z0-9_-]) after the quantified class) to prevent substring overmatching.
Eraminel01
force-pushed
the
fix/anthropic-api03-variable-length
branch
from
93ebdf4 to
7b598b0
Compare
santhsecurity
pushed a commit
that referenced
this pull request
May 29, 2026
Bump workspace version 0.5.35 -> 0.5.36 and the four pinned internal crate deps. Update README demo banner and the KEYHOG_VERSION / GitHub Action / pre-commit example pins to v0.5.36 so they match the tag. Ships: incremental never-cache-finding-bearing-files fix, hook fail-open, PR #7 anthropic-key shape + boundary tests, and the e2e lockdowns for hook/keyhogignore/UTF-16/incremental from the deep-coherence pass.
Owner
|
Hi eraminel01, Thank you for your contribution and welcome to the Team! Also, congrats on being the first external contributor to KeyHog. I will update with thanks for the contribution.md soon! -santh |
santhsecurity
pushed a commit
that referenced
this pull request
May 29, 2026
…h fixes
Clean-checkout CI was red for two structural reasons, both fixed here:
- keyhog-sources git source shared a gix 0.77 `Repository` (Send, NOT
Sync) across a Rayon `into_par_iter`, so a fresh
`cargo build -p keyhog-sources --features git` failed with 7
`RefCell<…> cannot be shared between threads` errors. Local builds only
passed on a cached artifact. Made the blob loop serial (correct; KH-58
re-parallelization via per-thread `gix::open` tracked as follow-up).
- Workspace-wide `cargo fmt` (the fmt gate was failing across many files).
Lean CI / embeddable build profile (`--no-default-features --features ci`):
- New cli `ci` feature: full 891-detector corpus + ML + entropy +
decode-through + multiline, but NO Hyperscan (kills the ~1.7s cold-start
-> ~0.85s and dropping), NO GPU/vyre drivers, NO live-verify, NO TUI,
NO Ghidra, NO git/web/cloud backends (filesystem + stdin cover `scan .`).
- keyhog-sources now declares `dashmap` directly (was only transitively
present, breaking reduced-feature builds) and sets `default-features =
false` at the WORKSPACE level so members' `default-features = false`
is actually honored (cargo ignores it otherwise).
- codewalk 0.2.5 `FileEntry { is_binary }` set in the explicit-include path.
Detection truth (all_tests: 32 -> 0 known failures):
- Boundary reassembly dedup is now detector-aware (offset+detector+hash),
so co-located findings from distinct detectors are no longer dropped on
the cross-chunk path (parity with in-chunk scan).
- Corpus de-duplication: shopify-access-token narrowed to shpca_ (admin/
storefront own shpat_/shpss_); cloudflare-workers WORKERS made mandatory
so plain cloudflare_api_token= belongs only to cloudflare-api-token.
- Concat reassembly: template-literal string interpolation (`${"x"}`),
Rust concat!(), and `${a}${b}` cross-line variable resolution.
- twilio-auth-token boundary test uses its required AccountSid companion;
heroku/homoglyph fixtures corrected to real-shape tokens.
Cold-start: HS DB cache moved from /tmp (wiped each reboot) to a persistent
XDG cache dir so the compile is paid once per machine, not once per boot.
Docs: README documents CI never needs a GPU + a concrete CI recipe;
CONTRIBUTING credits PR #7 (Edyard / @Eraminel01).
santhsecurity
pushed a commit
that referenced
this pull request
May 30, 2026
Applies the round 1 base64-protobuf FP-killer fixes from the investigator findings on the three owned files in this partition. shape_gates.rs::looks_like_standard_base64_blob: drop the strict +/-required clause; admit pure-alphanumeric mult-of-4 base64 in [40, 80] when alphabet diversity is >= 32 distinct alnum chars, and admit padded-no-punct base64 ending in = or ==. Covers cause #1 (25 FPs from pure-alnum) and cause #4 (6 FPs from padded-no-punct shapes). decode_structure.rs::looks_like_uniform_base64_blob: lower window floor from 60 to 44 (cause #7), add high-diversity admit (cause #2). Adds decoded_is_base64_blob helper for k8s data: double-base64 wrapper detection (cause #6) - decodes outer once and checks whether the inner bytes are themselves all-base64-alphabet of length >= 32. confidence/penalties.rs::apply_post_ml_penalties: wire decoded_is_base64_blob into the generic-* (non-named) arm with the same 0.02 slam as the uniform-blob and encoded-binary gates. Named detectors continue to bypass. Expected FP delta on base64-protobuf category: -52 (collapses the v32 surviving cluster). Each fix paired with positive truth + negative twin inline tests in the same file. Builds clean under both default and lean (no-default-features --features ml,entropy,decode,multiline) flags.
santhsecurity
pushed a commit
that referenced
this pull request
May 31, 2026
santhsecurity
pushed a commit
that referenced
this pull request
May 31, 2026
Bump workspace version 0.5.35 -> 0.5.36 and the four pinned internal crate deps. Update README demo banner and the KEYHOG_VERSION / GitHub Action / pre-commit example pins to v0.5.36 so they match the tag. Ships: incremental never-cache-finding-bearing-files fix, hook fail-open, PR #7 anthropic-key shape + boundary tests, and the e2e lockdowns for hook/keyhogignore/UTF-16/incremental from the deep-coherence pass.
santhsecurity
pushed a commit
that referenced
this pull request
May 31, 2026
…h fixes
Clean-checkout CI was red for two structural reasons, both fixed here:
- keyhog-sources git source shared a gix 0.77 `Repository` (Send, NOT
Sync) across a Rayon `into_par_iter`, so a fresh
`cargo build -p keyhog-sources --features git` failed with 7
`RefCell<…> cannot be shared between threads` errors. Local builds only
passed on a cached artifact. Made the blob loop serial (correct; KH-58
re-parallelization via per-thread `gix::open` tracked as follow-up).
- Workspace-wide `cargo fmt` (the fmt gate was failing across many files).
Lean CI / embeddable build profile (`--no-default-features --features ci`):
- New cli `ci` feature: full 891-detector corpus + ML + entropy +
decode-through + multiline, but NO Hyperscan (kills the ~1.7s cold-start
-> ~0.85s and dropping), NO GPU/vyre drivers, NO live-verify, NO TUI,
NO Ghidra, NO git/web/cloud backends (filesystem + stdin cover `scan .`).
- keyhog-sources now declares `dashmap` directly (was only transitively
present, breaking reduced-feature builds) and sets `default-features =
false` at the WORKSPACE level so members' `default-features = false`
is actually honored (cargo ignores it otherwise).
- codewalk 0.2.5 `FileEntry { is_binary }` set in the explicit-include path.
Detection truth (all_tests: 32 -> 0 known failures):
- Boundary reassembly dedup is now detector-aware (offset+detector+hash),
so co-located findings from distinct detectors are no longer dropped on
the cross-chunk path (parity with in-chunk scan).
- Corpus de-duplication: shopify-access-token narrowed to shpca_ (admin/
storefront own shpat_/shpss_); cloudflare-workers WORKERS made mandatory
so plain cloudflare_api_token= belongs only to cloudflare-api-token.
- Concat reassembly: template-literal string interpolation (`${"x"}`),
Rust concat!(), and `${a}${b}` cross-line variable resolution.
- twilio-auth-token boundary test uses its required AccountSid companion;
heroku/homoglyph fixtures corrected to real-shape tokens.
Cold-start: HS DB cache moved from /tmp (wiped each reboot) to a persistent
XDG cache dir so the compile is paid once per machine, not once per boot.
Docs: README documents CI never needs a GPU + a concrete CI recipe;
CONTRIBUTING credits PR #7 (Edyard / @Eraminel01).
santhsecurity
pushed a commit
that referenced
this pull request
May 31, 2026
Applies the round 1 base64-protobuf FP-killer fixes from the investigator findings on the three owned files in this partition. shape_gates.rs::looks_like_standard_base64_blob: drop the strict +/-required clause; admit pure-alphanumeric mult-of-4 base64 in [40, 80] when alphabet diversity is >= 32 distinct alnum chars, and admit padded-no-punct base64 ending in = or ==. Covers cause #1 (25 FPs from pure-alnum) and cause #4 (6 FPs from padded-no-punct shapes). decode_structure.rs::looks_like_uniform_base64_blob: lower window floor from 60 to 44 (cause #7), add high-diversity admit (cause #2). Adds decoded_is_base64_blob helper for k8s data: double-base64 wrapper detection (cause #6) - decodes outer once and checks whether the inner bytes are themselves all-base64-alphabet of length >= 32. confidence/penalties.rs::apply_post_ml_penalties: wire decoded_is_base64_blob into the generic-* (non-named) arm with the same 0.02 slam as the uniform-blob and encoded-binary gates. Named detectors continue to bypass. Expected FP delta on base64-protobuf category: -52 (collapses the v32 surviving cluster). Each fix paired with positive truth + negative twin inline tests in the same file. Builds clean under both default and lean (no-default-features --features ml,entropy,decode,multiline) flags.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Updates the Anthropic API key detector to match current observed sk-ant-api03- key shapes.
The previous regex required one fixed segment layout:
sk-ant-api03-[A-Za-z0-9_-]{93}-[A-Za-z0-9_-]{6}AA`n
That misses revoked real-world keys whose body length/segment layout differs. This keeps the distinctive sk-ant-api03- prefix and allows a variable URL-safe token body:
sk-ant-api03-[A-Za-z0-9_-]{80,120}`n
Also updates the Anthropic contract fixtures with revoked real key shapes.
Summary by CodeRabbit