test: add cross-layer, false positive, and mixed obfuscation adversarial payloads#11
Open
Ananya44444 wants to merge 3 commits intoc2siorg:mainfrom
Open
test: add cross-layer, false positive, and mixed obfuscation adversarial payloads#11Ananya44444 wants to merge 3 commits intoc2siorg:mainfrom
Ananya44444 wants to merge 3 commits intoc2siorg:mainfrom
Conversation
Initial set of 14 attack payloads organized by pipeline layer (prompt, context, normalization, memory/provenance). Each payload specifies which detection layer should catch it and the expected enforcement action, so we can track coverage gaps as modules get built. Ref: c2siorg#2
There was a problem hiding this comment.
Pull request overview
Extends the adversarial payload taxonomy under tests/adversarial/ to broaden pipeline-validation coverage, adding additional payload sets intended to exercise cross-layer decoding/normalization chains, precision (false positives), and mixed obfuscation patterns.
Changes:
- Adds new prompt-, context-, normalization-evasion-, memory-, and cross-layer payload JSON libraries.
- Introduces a benign “false positive” payload case to validate non-overblocking behavior.
- Adds/updates the adversarial taxonomy README describing schema and coverage.
Reviewed changes
Copilot reviewed 6 out of 6 changed files in this pull request and generated 6 comments.
Show a summary per file
| File | Description |
|---|---|
| tests/adversarial/payloads/prompt_layer.json | Prompt-injection payloads, including multi-turn and a benign false-positive case |
| tests/adversarial/payloads/context_layer.json | Context-layer payloads (RAG/tool-output reinjection, context flooding) |
| tests/adversarial/payloads/normalization_evasion.json | Encoding/obfuscation payloads (homoglyph, base64, zero-width, leetspeak, mixed) |
| tests/adversarial/payloads/memory_layer.json | Memory/provenance attack payloads (poisoning, spoofing, replay) |
| tests/adversarial/payloads/cross_layer.json | Cross-layer payloads intended to become malicious post decoding/normalization |
| tests/adversarial/README.md | Documentation for payload organization, schema, and coverage matrix |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
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
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.
Extends the adversarial payload library from #5 with additional real-world attack patterns.
Additions
Cross-layer payloads where malicious intent emerges post-normalization/decoding
False positive test case to evaluate precision and avoid overblocking
Mixed obfuscation payload combining zero-width characters and homoglyphs
Motivation
The current taxonomy focuses on layer-isolated behavior. These additions help validate:
Correct stage ordering (Normalize → Scan)
Robustness against encoding-based evasion chains
Precision of detection systems (false positives)
Built on top of #5