[Feature Request] Memory Poisoning Protection for Haystack via OWASP Agent Memory Guard

[vgudur-dev]

Problem

Haystack pipelines using memory stores and RAG are vulnerable to memory poisoning attacks — adversarial inputs in document stores or conversation memory can cause pipelines to produce corrupted outputs. OWASP identifies this as a top risk for LLM applications.

Proposed Solution

OWASP Agent Memory Guard (AMG) is an open-source Python library that wraps any memory store as a transparent security layer:

• pip install agent-memory-guard
• Scans every memory write for prompt injection, PII leakage, and tampering
• 92.5% detection rate on AgentThreatBench benchmark
• Works with any memory backend

Links

• GitHub: https://github.com/OWASP/www-project-agent-memory-guard
• PyPI: https://pypi.org/project/agent-memory-guard/
• Benchmark: https://pypi.org/project/agent-threat-bench/

Would the Haystack team consider integrating AMG as a security layer? Happy to contribute a PR.

[julian-risch]

@vgudur-dev Thank you for opening an issue brining up this idea! We recommend to publish and release this integration yourself using our repository template if you believe it benefits a larger group of users. https://github.com/deepset-ai/custom-component
You could include an example with mem0 or cognee as memory backends that have a cognee integration.