feat(ai-red-teaming): honeytoken tool (mint/use canaries) - #129
Open
rdheekonda wants to merge 1 commit into
Open
feat(ai-red-teaming): honeytoken tool (mint/use canaries)#129rdheekonda wants to merge 1 commit into
rdheekonda wants to merge 1 commit into
Conversation
… language) Adds the deferred interactive honeytoken tool to the ai-red-teaming capability: - mint_honeytoken(kind=exfil|rce, location) -> an inert, self-labeling canary plus a ready-to-plant injection string. - check_honeytoken_leaked(canary_value, agent_output) -> representation-invariant verdict (base64/hex/url/unicode canonicalized) proving exfil/execution. Thin TUI-facing front end over the SDK harness (dreadnode.airt.honeytoken); the same canaries/scorers back the OWASP-ASI suite runner. Tests skip on an SDK build without the honeytoken module and pass 7/7 on one that has it. Note: requires the SDK honeytoken harness (dreadnode-tiger #2181) to be released; functional at runtime once the capability's SDK pin includes it.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Adds the deferred honeytoken tool to the
ai-red-teamingcapability — the interactive/TUI front end to the SDK honeytoken harness (the follow-up called out in dreadnode-tiger #2181):mint_honeytoken(kind=exfil|rce, location)→ an inert, self-labeling canary + a ready-to-plant injection string (exfil forwarder / RCE proof-of-execution beacon).check_honeytoken_leaked(canary_value, agent_output)→ representation-invariant verdict (base64/hex/url/unicode canonicalized first) —LEAKEDis hard evidence the token crossed the boundary.Effect-based, not string-matching; the canary is a benign never-real token, so leaking it exposes nothing. Same canaries/scorers back the OWASP-ASI suite runner.
Dependency
Requires the SDK honeytoken harness
dreadnode.airt.honeytoken(dreadnode-tiger #2181). The testspytest.importorskipit, so CI is green on the current SDK pin (they skip) and run once the pin includes it. Bump the capability's SDK pin after #2181 releases to make the tool functional at runtime.Tests
tests/test_honeytoken_tool.py— mint (exfil/rce/invalid), plaintext + base64-obfuscated detection, clean-output negative, missing-value guard. Skips on an SDK without honeytoken; 7 passed against one that has it.Test plan