by depalmar
Provides a vendor‑neutral collection of Sigma, YARA, and Suricata rules for detecting compromises of LLM applications, MCP servers, AI supply chains, and related insider‑threat scenarios.
AI ML DFIR Detection Pack delivers open‑source detection signatures covering a wide range of AI/ML attack vectors—including prompt injection, MCP configuration poisoning, model supply‑chain exploits, AI‑infrastructure abuse, RAG/vector‑DB attacks, and runtime AI‑malware. All rules are written in open standards (Sigma .YML, YARA .YAR, Suricata .RULES) and can be deployed on any modern SIEM, EDR, or network IDS.
https://depalmar.github.io/ai-dfir-toolkit/ for rule catalogs and mappings.pip install sigma-cli pysigma-backend-elasticsearch
sigma convert -t lucene --without-pipeline ai-dfir-toolkit/**/*.yml > ai-dfir-toolkit.lucene
yara -r ai-dfir-toolkit/03-model-supply-chain/*.yar /path/to/models/
cp ai-dfir-toolkit/**/*.rules /etc/suricata/rules/
echo 'rule-files: [ai-dfir.rules]' >> /etc/suricata/suricata.yaml
suricata -T -c /etc/suricata/suricata.yaml
systemctl reload suricata
cd tests && ./validate.sh
Expected outcome: 15 passes, 0 failures.MAPPINGS.md links each rule to MITRE ATLAS, MITRE ATT&CK, CISA KEV, and OWASP LLM/Agentic 2026 lists.Q: Do I need proprietary software to run these rules?
A: No. All signatures are in open formats and can be converted to any SIEM or IDS using free tools like pySigma, yara, and Suricata.
Q: How often are the rules updated? A: The GitHub Actions workflow rebuilds the documentation site on every push; rule updates are merged via pull requests and undergo CI validation.
Q: Can I contribute new signatures? A: Yes. Follow the contribution guide: reference an ATLAS/ATT&CK technique or CVE, include test data, document false positives, and name files using snake_case.
Q: What platforms are supported for the artifact catalog? A: The catalog feeds into ForensicArtifacts, KAPE, Velociraptor, and generic CSV/JSON consumers.
Q: Is there a recommended severity tuning process?
A: Deploy the rules to a test environment for 7 days, triage hits, adjust selection and filter blocks, then promote rules with level: high or above to production.
Open-source detection signatures for AI/ML attacks and breaches.
Supported by SANS Institute. The project stays independent and vendor-neutral; see the disclaimer below.
A vendor-neutral collection of Sigma, YARA, and Suricata rules for detecting compromise of LLM applications, MCP servers, ML supply chains, AI infrastructure, AI-powered insider threats, and RAG/vector database attacks.
Browse it at depalmar.github.io/ai-dfir-toolkit — the artifact catalog, the detection rules, the ATLAS/OWASP mappings, the case studies, and the investigation guide, rebuilt from this repository on every push so the page cannot drift from the content.
Start from the alert. Name the tool the alert mentioned, say where the AI ran and what role it played, and get its traces sorted by how fast each one disappears.

Collect in the right order. Pick rows into a plan, group it by volatility, and export it for the collector your team already runs. Credentials are hashed, not copied.

Full-size video: start from the alert · collection plan. How they were recorded: docs/media/.
For the companion investigation guide with attack background, forensic artifacts, Mermaid attack-chain diagrams, and hands-on investigation procedures, see docs/ai-dfir-investigation-guide.md.
License: Apache 2.0.
This is an independent personal project. It is not affiliated with, endorsed by, or produced on behalf of any employer. SANS Institute supports the project and has cleared its logo for use here; that support does not change the project's independence, and the content remains the maintainer's own. All research is based on public sources — published CVEs, vendor advisories, academic papers, and vendor-neutral security research. Contributions are welcome from the community.
Most existing detection content is either locked behind vendor SIEMs or scattered across blog posts. AI/ML attacks need detection coverage that spans:
This pack uses open standards only so the rules can be deployed in any modern detection stack via open Sigma/YARA/Suricata tooling.
ai-dfir-toolkit/
├── 01-llm-prompt-injection/ # Prompt injection, jailbreaks, indirect injection
├── 02-mcp-attacks/ # MCP tool poisoning, config tampering, rug pulls
├── 03-model-supply-chain/ # Pickle exploits, HuggingFace, dependency confusion
├── 04-ai-infrastructure/ # ShadowRay, Triton, MLflow, GPU abuse
├── 05-copilot-assistant-abuse/ # M365 Copilot, GitHub Copilot, Claude, Cursor
├── 06-rag-vector-db/ # Vector DB exposure, RAG poisoning
├── 07-runtime-ai-malware/ # Malware calling an LLM API at runtime
├── 08-agentic-orchestration/ # Agent-as-operator, AI service API as C2
├── 09-agent-memory-forensics/ # Memory as persistence, context poisoning
├── collectors/ # Read-only acquisition with chain of custody
├── playbooks/ # CACAO v2.0 response playbooks
├── artifacts/ # Machine-readable AI agent artifact catalog
├── skills/ # Agent skills for maintaining the catalog
├── docs/ # Investigation guide
├── tests/ # Sample events / test files
├── MAPPINGS.md # ATLAS + OWASP cross-reference
└── README.md
Each category directory contains a README.md describing the threats covered and rule files in their native format (.yml for Sigma, .yar for YARA, .rules for Suricata).
| Format | Use Case | Where it Deploys |
|---|---|---|
Sigma (.yml) |
Generic log-based detection | Any SIEM via pySigma backends |
YARA (.yar) |
File / memory artifacts | EDR platforms, malware analysis, file scanning pipelines |
Suricata (.rules) |
Network traffic | Suricata, Snort (compatible subset), Zeek (via translation) |
All rules use open formats. No vendor-specific query languages, no proprietary field schemas. Convert to your platform using pySigma backends.
pip install sigma-cli pysigma-backend-elasticsearch
sigma convert -t lucene --without-pipeline ai-dfir-toolkit/**/*.yml > ai-dfir-toolkit.lucene
The rules are vendor-neutral Sigma — see the pySigma backends list to convert to any other SIEM query language.
Case sensitivity: the keyword-based rules (prompt injection, jailbreak, system-prompt extraction, etc.) follow the Sigma convention that
containsmatching is case-insensitive — attacker text varies in case, so the rules are written in lowercase and rely on the backend to fold case. On Elastic, ensure the matched content fields are mapped as analyzedtext(the default for string fields), notkeyword: akeywordmapping matches case-sensitively and will miss capitalized input. Do not try to force this with there|imodifier — Lucene's regex engine does not support the(?i)flag and the resulting query is invalid.
# Scan a model directory
yara -r ai-dfir-toolkit/03-model-supply-chain/*.yar /path/to/models/
# Scan an MCP config
yara ai-dfir-toolkit/02-mcp-attacks/mcp_tool_poisoning.yar \
~/Library/Application\ Support/Claude/claude_desktop_config.json
cp ai-dfir-toolkit/**/*.rules /etc/suricata/rules/
echo 'rule-files: [ai-dfir.rules]' >> /etc/suricata/suricata.yaml
suricata -T -c /etc/suricata/suricata.yaml # validate
systemctl reload suricata
artifacts/ is a machine-readable catalog of the endpoint
artifacts left by AI coding agents, local LLM runtimes, agentic workflow
engines, and Model Context Protocol components — install paths, credential
storage, MCP configs, listening ports, process relationships, and registry keys,
each rated by forensic value and sourcing confidence, with collection priorities
for IR triage.
Eleven tool categories are covered: local LLM runtimes, coding agents and IDEs,
agent frameworks, coding-agent CLIs, workflow engines, browser agents,
computer-use agents, MCP hosts, cloud agents, code-completion tools, and
code-execution agents. Alongside them sit documented incident case studies, each
carrying its own provenance and recording analyst disagreement in a contested
field rather than resolving it.
Ships vendor-neutral Sigma detection rules and an osquery inventory pack, plus generated feeds in four downstream formats:
| Export | Consumed by |
|---|---|
| ForensicArtifacts | Plaso, GRR, Timesketch |
| KAPE targets | KAPE triage collections |
| Velociraptor artifacts | Velociraptor hunts |
| CSV / JSON | anything else — see artifacts/docs/api/ |
On top of the catalog sits a threat-intel layer:
Technique and vulnerability ids are checked against pinned releases of MITRE
ATLAS, MITRE ATT&CK, the CISA KEV catalog, and the OWASP LLM and Agentic 2026
lists. The pins are recorded in artifacts/schema/ and a
weekly workflow proposes updates.
Each profile is derived from the cases and reported sightings that cite it. It gives the techniques the actor uses, the rule in this repository that detects each one, the catalogued tools the actor abused, and which of those tools' artifacts to collect first. The site adds:
Feeds: intel.json, a
STIX 2.1 bundle for OpenCTI or
MISP, and Maltego import CSVs. Two scheduled
workflows keep it current:
intel-candidate issues for a person to
researchFor current counts — catalogued tools, documented artifacts, credential
locations, MCP config paths, case studies, and the risk and confidence
breakdowns — see the Contents table in artifacts/README.md.
That table is generated by scripts/readme_counts.py and gated in CI, so it is
the one place those numbers are guaranteed current.
76 rule files containing 171 individual signatures — nine attack-class categories plus a cross-tool endpoint set:
| Category | Files | Signatures | ATLAS Techniques | OWASP LLM 2026 | OWASP Agentic |
|---|---|---|---|---|---|
| LLM Prompt Injection | 8 | 10 | T0051, T0054, T0029 | LLM01, LLM02, LLM06, LLM08 | ASI01 |
| MCP Attacks | 5 | 14 | T0010, T0115.002, T0110, T0086 | LLM02, LLM03, LLM04 | ASI02, ASI03, ASI04, ASI05 |
| Model Supply Chain | 8 | 25 | T0010, T0018, T0020 | LLM04 | ASI03, ASI04, ASI05 |
| AI Infrastructure | 9 | 31 | T0011, T0049, T0029 | LLM06 | — |
| Copilot/Assistant Abuse | 8 | 19 | T0086, T0024 | LLM02, LLM03, LLM04 | ASI01, ASI02, ASI03, ASI04, ASI05 |
| RAG / Vector DB | 5 | 19 | T0020 | LLM01, LLM02, LLM09 | ASI01, ASI06 |
| Runtime AI-Malware | 8 | 16 | T0096, T0086 | LLM01, LLM03 | — |
| Agentic Orchestration & AI-Service C2 | 8 | 17 | T0096, T0086, T0054, T0012, T0055, T0116 | LLM03 | ASI02 |
| Agent Memory & Context Poisoning | 2 | 5 | T0080, T0080.000, T0086 | LLM01, LLM02, LLM03 | ASI06 |
| Endpoint (cross-tool) | 15 | 15 | T0053, T0081, T0083, T0098, T0010.001 | LLM02, LLM03, LLM04, LLM06 | ASI02, ASI03, ASI04, ASI05 |
| Total | 76 | 171 |
The endpoint set lives in artifacts/detections/ and is
scoped to agent behaviour on a host rather than to one attack class, so it applies
across every tool in the artifact catalog. An osquery pack there answers the
inventory question ("which hosts have this") that the Sigma rules cannot.
Signature count includes multi-document Sigma YAML, multiple rule blocks inside a single YARA file, and multiple alert lines inside a single Suricata .rules file, except one flowbits:noalert helper that only sets state for the rule after it. One file often covers several related variants.
See MAPPINGS.md for per-rule mappings.
These rules are written to err toward signal over noise, but every environment is different. Each rule includes:
falsepositives: section listing known FP scenarioslevel: field (low / medium / high / critical) — start with high+ in productionRecommended rollout:
selection and filter blocksThe tests/ directory contains sample artifacts (malicious and benign) for validating rule correctness. Run:
cd tests/
./validate.sh
Expected result: 15 passes, 0 failures.
The suite hard-exits if the yara binary is missing rather than reporting a
clean run — absent tooling would otherwise score every "expect no matches"
assertion as a pass and the suite would look healthy while testing nothing.
PRs welcome. Rule submission requirements:
tests/ — sample log line, file, or PCAPattack.atlas.txxxx in Sigma rulesSigma format: follow the Sigma specification.
Apache License 2.0 — see LICENSE.
You are free to use, modify, and redistribute these rules in commercial and non-commercial settings. Attribution appreciated but not required.
Catalog data under artifacts/catalog/, artifacts/case-studies/, and
artifacts/docs/api/ is licensed CC BY 4.0; see
artifacts/LICENSE-DATA. All other content,
including the scripts and schema under artifacts/, is under the repository's
Apache-2.0 licence.
Maintainer: Raymond DePalma — independent security researcher. This is a personal project and is not affiliated with any employer. SANS Institute's support is described in the disclaimer.
Please log in to share your review and rating for this MCP.
Explore related MCPs that share similar capabilities and solve comparable challenges
by chaitin
A self‑hosted web application firewall and reverse proxy that protects web applications from attacks and exploits by filtering, monitoring, and blocking malicious HTTP/S traffic.
by beenuar
Ingests security telemetry, normalizes it, runs a large library of detection rules, correlates alerts, and leverages AI agents for triage and investigation while recording every prompt, tool call, and decision in an immutable ledger.
by snyk
Scans installed AI agent components, MCP servers, and skill files for prompt‑injection, tool poisoning, toxic flows, hard‑coded secrets and other supply‑chain risks.
by OpenOSINT
Provides an AI‑driven OSINT workflow that lets users query a natural‑language REPL, CLI, web UI, or MCP server, automatically selecting and chaining 18 reconnaissance tools to collect, pivot, verify, and report public‑source intelligence.
by 2akouwu
Provides deterministic verification of AI‑generated claims against binary artifacts, ensuring every structural or behavioral assertion is grounded in the actual file and preserving verified state across context resets.
by hashgraph-online
Open-source antivirus for AI agents that blocks risky tools, secret exposure, prompt injection, malicious packages, MCP servers, plugins, and skills at runtime.
by safedep
Provides enterprise‑grade open source software supply chain security by scanning source code, dependencies, containers and SBOMs, detecting vulnerabilities and malicious packages, and enforcing policy as code.
by tufantunc
Provides controlled SSH access for LLM agents with command classification, policy‑based authorization, human‑in‑the‑loop approval, and immutable audit logging.
by semgrep
Offers an MCP server that lets LLMs, agents, and IDEs run Semgrep scans to detect security vulnerabilities in source code.