feat: semia-to-ave crosswalk - #190
Merged
Merged
Conversation
…#186) Co-authored-by: chaksaray <15962335+chaksaray@users.noreply.github.com>
Mechanism-level verification against Semia's real action/effect structure, not label matching. Note field states directly whether the constraint-guided synthesis approach generalizes onto AVE's classes, per the open question from issue #36.
# Conflicts: # dist/ave-records-latest.manifest.json
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
First-pass crosswalk per berabuddies/Semia#36. 16 verified matches (14 distinct AVE ids) across 9 of Semia's 11 detector rules, mechanism-level, not label-based. Note field answers the SDL-generalization question raised in the original outreach.
What was actually read
Not just the paper's act_sign/c_sign example.
packages/semia-core/src/semia_core/schema.py(the complete SDL fact vocabulary, 16 effects/4 triggers/5 gates/5 doc claims/7 value kinds/6 call codes) andpackages/semia-core/src/semia_core/rules/sdl/skill_dl_static_analysis.dl(the 11 reallabel_*Datalog detector rules Semia's owndetector.pyreads back out as findings).The generalization answer (not left implicit in the mapping count)
Partial, in a specific and citable way: clean where a Semia rule terminates in a specific effect AVE also names explicitly (crypto_sign, code_eval, proc_exec chains), genuinely distinct where AVE organizes by attack narrative and Semia organizes by dataflow destination. Three concrete examples in the crosswalk's own
notefield:label_unsanitized_context_ingestionkeyword-sweeps ~2 dozen AVE prompt-injection records but only structurally verifies against ones naming a specific privileged sink (2 of them here).label_hardcoded_c2_communication's name promises a match its logic doesn't deliver: it requires an unresolvable target, while AVE-2026-00073's mechanism is a fully resolved, literal bad destination. Left unmapped rather than forced.label_dormant_malicious_payload's narrative but the mechanism spans multiple agents' memories, a dimension Semia's single-skill()-per-pass schema has no representation for at all. Left unmapped.Also surfaced: AVE-2026-00003 is now matched by 4 independent tools (cfgaudit, nova-proximity, Ramparts, and now Semia) plus 2 of Semia's own internal rules converging on it independently.
Validation
python3 scripts/validate_crosswalks.py-> 7/7 valid, this crosswalk the only one with zero warnings (both source commit and target commit pinned, sorecord_count/static_record_countare re-derivable, unlike the pre-existing crosswalks).Base branch note
Targets
developper this repo's established convention, not the task brief's literal--base main.🤖 Generated with Claude Code