<!-- Aegis skill mitre-atlas v1.1.1 (osk 1.1). Generated file: edit the source skill, not this. -->

> Intel as of 2026-08-22.

# MITRE ATLAS Skill

## Scope and authorization

- Purpose: Threat-model AI and ML systems with MITRE ATLAS and plan countermeasures against attacks on them.
- Use only on systems you own or are explicitly authorized to assess.
- Out of scope: poisoning, evading or extracting models you do not own, and producing working attacks against deployed third-party AI systems.
- No network or API calls by the assistant: it never makes network requests, calls APIs or contacts outside services. Commands here are for a human operator to review and run. Redact secrets, credentials, personal data and samples before sharing anything.


## Purpose

Assess the attack surface of AI/ML systems using the MITRE ATLAS framework and
design appropriate deception or detection assets. ATLAS is ATT&CK's counterpart
for adversarial machine learning — it maps how adversaries attack AI systems specifically.

## Phase Map

```
Phase 0 → Scope the AI/ML system and identify components in scope
Phase 1 → ATLAS tactics orientation      [read: references/atlas-tactics-overview.md]
Phase 2 → ML attack surface assessment   [read: references/ml-attack-surface-guide.md]
Phase 3 → Output: ML threat model + deception asset recommendations per zone
```

## Phase 0 — Scope

Establish:
1. **ML system components in scope**: training pipeline, model registry, inference API, feature store, credentials
2. **Model type**: custom-trained / fine-tuned / API-only (third-party model)
3. **Data sensitivity**: what training data contains, what inference inputs reveal
4. **Adversary goal**: IP theft (model weights) / integrity attack (poisoning) / availability / evasion

Proceed to Phase 1 once scope is clear.

## Output Format

For each ML zone in scope, produce:

| ML Zone | ATLAS Technique | Deception Asset | Signal Routing |
|---------|----------------|-----------------|---------------|
| ML Credentials | AML.T0012 Valid Accounts | Honeytoken API keys | CloudTrail → SIEM |
| Training Pipeline | AML.T0020 Training Data Poisoning | Canary training samples | Model output monitor |

## ATLAS Technique Index — AML.T Series

### Reconnaissance Techniques (AML.TA0002 Reconnaissance)
| Technique | ID | ATT&CK Analogue | Countermeasure |
|-----------|-----|-----------------|----------------|
| Search for Victim's Publicly Available Research Materials | AML.T0000 Search Open Technical Databases | T1593 | Monitor academic publication of model details |
| Search Victim's ML Model via Inference API | AML.T0040 AI Model Inference API Access | T1595.002 | Rate-limit inference API; canary queries on hosted endpoints |
| Acquire Public ML Artifacts | AML.T0002 Acquire Public AI Artifacts | T1588.002 | Monitor Hugging Face / PyPI for model clones |
| Search GitHub/GitLab for ML Code | AML.T0095 Search Open Websites/Domains | T1593.003 | Secret scanning on git repos containing keys and model weights |
| Develop ML Attack Capabilities | AML.T0017 Develop Capabilities | T1587 | N/A — adversary development |
| Obtain ML Attack Capabilities | AML.T0016 Obtain Capabilities | T1588 | Monitor darkweb for ML attack toolkits |

### AI Model Access (AML.TA0000 AI Model Access)
| Technique | ID | Description |
|-----------|----|-------------|
| Valid Accounts for ML Services | AML.T0012 Valid Accounts | Credential theft against Vertex AI, SageMaker, Hugging Face — honeytoken API keys in vaults |
| ML Service Tokens | AML.T0055 Unsecured Credentials | Bearer token theft from API endpoints |
| Inference API Access via Proxy | AML.T0040 AI Model Inference API Access | Reverse proxy as canary — logs all ML API queries |
| Publish Backdoored ML Model | AML.T0010.003 AI Supply Chain Compromise: Model | Supply chain attack via PyPI / Conda — internal PyPI mirror |

### Training Data Attacks (AML.TA0006 Persistence)
| Technique | ID | Real-World Example | Deception Response |
|-----------|----|--------------------|-------------------|
| Poison Training Data | AML.T0020 Training Data Poisoning | Adversary injects mislabelled samples into the training pipeline; T1195 Supply Chain Compromise analogue | Canary samples: uniquely identifiable data points that fire if extracted |
| Backdoor ML Model | AML.T0018 Manipulate AI Model | Insert trigger pattern into model weights stored on an NFS share; T1554 Compromise Client Software Binary analogue | Honeypot model registry version with access-logged weights |
| Craft Adversarial Training Data | AML.T0043 Craft Adversarial Data | Manipulate feature values in data streams feeding training | Canary feature values with known signatures |

### Inference Attacks (AML.TA0009 Collection)
| Technique | ID | Attack Description | Detection |
|-----------|----|--------------------|-----------|
| Model Inversion Attack | AML.T0024.001 Invert AI Model | Adversary queries inference API repeatedly to reconstruct training data (membership inference is AML.T0024.000 Infer Training Data Membership). T1119 Automated Collection analogue | Rate-limit + canary query detection |
| Model Evasion | AML.T0015 Evade AI Model | Adversarial examples crafted to bypass detection models (FGSM, PGD attacks against the ML pipeline). T1562 Impair Defenses analogue | Shadow model with intentional evasion logging |
| Model Extraction | AML.T0024.002 Extract AI Model | Query ML inference API exhaustively to reconstruct model weights. T1119 + T1048 analogue | Honeypot model endpoint — query patterns fire alert |

### Exfiltration (AML.TA0010 Exfiltration)
| Technique | ID | Description |
|-----------|----|-------------|
| Exfiltrate via Cyber Means | AML.T0025 Exfiltration via Cyber Means | Steal model IP over C2 channel from compromised training host; T1041 Exfiltration over C2 |
| Exfiltrate ML Model | AML.T0025 Exfiltration via Cyber Means | Copy model weights from registry to external storage via T1537 Transfer Data to Cloud Account |

## ML Attack Examples — Worked Scenarios

### Data Poisoning Attack (AML.T0020 Training Data Poisoning)
**Target:** Image classification model for fraud detection  
**Vector:** Adversary with write access to training data bucket (T1530 Data from Cloud Storage Object)  
**Tools:** ML data pipeline, Python TensorFlow training loop  
**Attack:** 3% of training images replaced with adversarially perturbed samples labelled as benign  
**Countermeasures:** Canary samples (known-hash images) → if model output changes on canaries, poisoning confirmed  
**ATT&CK mapping:** T1195.001 → T1565.001 → T1486 (integrity attack)

### Model Inversion Attack (AML.T0024.001 Invert AI Model)
**Target:** Healthcare NLP model exposed via TLS-terminated REST API behind a gateway  
**Tools:** Adversary uses Python script + Hugging Face transformers to craft membership inference queries  
**Attack:** 50,000 queries to inference API reconstruct 73% of training data patients' diagnoses  
**Countermeasures:**  
- Canary query registration: known inputs that fire if submitted  
- Gateway rate-limiting + anomaly detection on query patterns  
- Honeytoken patient records in training set — appear in inversion output to confirm attack  
**ATT&CK mapping:** T1119 → T1040 → T1048

### Model Evasion (AML.T0015 Evade AI Model)
**Target:** Network intrusion detection model (streaming input feed)  
**Tools:** Foolbox adversarial attack library, FGSM/PGD perturbation generation  
**Attack:** Network packets perturbed by ε=0.01 to evade detection while maintaining attack effectiveness  
**Countermeasures:**  
- Shadow model trained without adversarial robustness — detects when main model and shadow diverge  
- Signed audit log of model decisions with input hashes  
**ATT&CK mapping:** T1562.001 → T1562.006 → T1036

## Phases

Load each phase only when the workflow reaches it:

- [Atlas tactics overview](references/atlas-tactics-overview.md)
- [Ml attack surface guide](references/ml-attack-surface-guide.md)
- [Ml threat model](references/ml-threat-model.md)
- [Ml defense controls](references/ml-defense-controls.md)
- [Ml incident response](references/ml-incident-response.md)
- [Live threat intel](references/live-threat-intel.md) — auto-updated weekly

---

## Phase: atlas-tactics-overview

# MITRE ATLAS Tactics Overview

ATLAS (Adversarial Threat Landscape for AI Systems) covers attacks against
ML/AI systems. Use this reference during Phase 1 to map the AI attack surface.

## ATLAS Tactics

| ID | Tactic | What the adversary targets |
|----|--------|---------------------------|
| AML.TA0000 | AI Model Access | Obtain query or training access to model |
| AML.TA0001 | AI Attack Adaptation | Adapt capabilities, methods, knowledge or objectives into attack-ready form for the target or current conditions |
| AML.TA0002 | Reconnaissance | Gather info about target ML system |
| AML.TA0003 | Resource Development | Acquire tools, infrastructure, training data |
| AML.TA0004 | Initial Access | Gain foothold via ML API, supply chain, or data pipeline |
| AML.TA0005 | Execution | Run malicious inputs against the model |
| AML.TA0006 | Persistence | Backdoor model weights, poisoned training sets |
| AML.TA0007 | Defense Evasion | Craft adversarial inputs that evade detection |
| AML.TA0008 | Discovery | Probe model capabilities, architecture, training data |
| AML.TA0009 | Collection | Extract training data, model weights via inversion |
| AML.TA0010 | Exfiltration | Steal model intellectual property |
| AML.TA0011 | Impact | Degrade model accuracy, manipulate outputs |

## High-Priority ATLAS Techniques

| ID | Technique | ATT&CK Analogue |
|----|-----------|----------------|
| AML.T0012 | Valid Accounts (valid ML service credentials) | T1078 Valid Accounts |
| AML.T0015 | Evade AI Model (evasion via adversarial examples) | T1562 Impair Defenses |
| AML.T0016 | Obtain Capabilities: ML Artifacts | T1588 Obtain Capabilities |
| AML.T0020 | Training Data Poisoning (poison training data) | T1195 Supply Chain Compromise |
| AML.T0024.001 | Invert AI Model (ML model inversion attack) | Data theft / IP exfiltration |
| AML.T0040 | AI Model Inference API Access | T1133 External Remote Services |
| AML.T0043 | Craft Adversarial Data | Execution / Impact |
| AML.T0115.000 | Publish Poisoned AI Artifacts: Datasets (publish poisoned datasets) | Supply chain compromise |

Full case studies: https://atlas.mitre.org/studies

---

## Phase: ml-attack-surface-guide

# ML Attack Surface Guide

Use this reference during Phase 2 to characterise the AI/ML attack surface
before designing deception or detection assets.

## ML System Zones

| Zone | Components | Deception Primitives |
|------|------------|---------------------|
| Training Data Pipeline | Data ingestion, labelling, storage | Canary samples in training sets — fire if model reproduces them |
| Model Registry | Stored weights, versioned checkpoints | Honeypot model version — access logged |
| Inference API | Public or internal API endpoint | Canary queries — specific inputs that trigger logging |
| Feature Store | Feature engineering, real-time features | Fake features — access fires alert |
| ML Credentials | API keys, service accounts for ML platforms | Honeytoken API keys for Hugging Face, OpenAI, Vertex AI |
| Model Supply Chain | Pre-trained models, fine-tuning datasets | Canary weights — detectable fingerprints in model output |

## Deception for ML Systems

**Canary training samples:** Inject a small number of uniquely identifiable
data points into training sets. If a model inversion attack extracts training
data, the canary samples appear in the extracted set — confirming the attack.

**Honeypot model endpoints:** Deploy a slightly degraded model version with
access logging. Any API query to the honeypot endpoint is a confirmed reconnaissance signal.

**Honeytoken ML credentials:** Fake API keys for ML platforms (Hugging Face,
OpenAI, Vertex AI, AWS SageMaker). Any use of these keys fires immediately —
the adversary has stolen credentials from the environment.

**Canary inference queries:** Register specific input strings. If these strings
appear in inference logs, the model is being probed for membership inference.

## Assessment Questions

1. Does the organisation train custom models? → risk of training data poisoning
2. Are model weights stored and versioned? → risk of weight theft
3. Is there a public or partner-facing inference API? → risk of model inversion
4. Are ML platform credentials in code repos or CI/CD? → honeytoken opportunity
5. Does the organisation use pre-trained models from public sources? → supply chain risk

---

## Phase: ml-threat-model

# ML Threat Model — Reference

Use during Phase 3 to produce a structured threat model for the AI/ML system in scope,
mapping each identified threat to an ATLAS technique and designing a countermeasure.

## ML System Data Flow Diagram (DFD) Components

| DFD Component | Examples | Trust Level | Primary ATLAS Threats |
|---------------|----------|-------------|----------------------|
| External data sources | Web scraping pipeline, vendor feeds, S3 buckets | Untrusted | AML.T0020 Training Data Poisoning via external feed |
| Training pipeline | Apache Spark, Kubeflow pipelines, MLflow | Semi-trusted | AML.T0020 Training Data Poisoning |
| Model registry | MLflow artifacts, Hugging Face Hub, AWS S3 | Trusted | AML.T0018 Manipulate AI Model (backdoor ML model) |
| Inference API | nginx-fronted REST endpoint, gRPC service | Public/Semi-public | AML.T0024.001 Invert AI Model, AML.T0024.002 Extract AI Model |
| Feature store | Feast, Tecton, Redis feature cache | Trusted | AML.T0043 Craft Adversarial Data |
| ML credentials | API keys for Vertex AI, SageMaker, Hugging Face, OpenAI | Highly sensitive | AML.T0012 Valid Accounts (valid ML service credentials) |
| CI/CD pipeline | GitHub Actions running training jobs on linux runners | Semi-trusted | AML.T0010.001 AI Supply Chain Compromise: AI Software — poisoned dependency |

## STRIDE Applied to ML Systems

| STRIDE Category | ML-Specific Threat | ATLAS Technique | ATT&CK Analogue |
|-----------------|-------------------|-----------------|-----------------|
| Spoofing | Adversary impersonates ML API consumer to extract model | AML.T0012 | T1078 Valid Accounts |
| Tampering | Adversary modifies training data in upstream S3 bucket (apache S3-compatible) | AML.T0020 Training Data Poisoning | T1565 Data Manipulation |
| Repudiation | No audit log on model registry — weight changes undetected | AML.T0018 Manipulate AI Model | T1562 Impair Defenses |
| Information Disclosure | Model inversion attack extracts training data from openssl-wrapped API | AML.T0024.001 Invert AI Model | T1213 Data from Information Repositories |
| Denial of Service | Flooding inference API with adversarial inputs saturates GPU resources | AML.T0029 Denial of AI Service | T1499 Endpoint DoS |
| Elevation of Privilege | Compromised feature store allows adversary to inject arbitrary features | AML.T0043 Craft Adversarial Data | T1548 Abuse Elevation |

## ML Threat Register Template

| Threat ID | Component | ATLAS Technique | STRIDE | Severity | Likelihood | Countermeasure | Owner |
|-----------|-----------|----------------|--------|----------|------------|----------------|-------|
| ML-T001 | Training data S3 bucket | AML.T0020 Training Data Poisoning | Tampering | Critical | Medium | Canary samples + hash verification | MLOps |
| ML-T002 | Model registry | AML.T0018 Manipulate AI Model | Tampering | High | Low | Signed model artifacts; immutable registry | MLOps |
| ML-T003 | Inference API (nginx) | AML.T0024.002 Extract AI Model | Info Disclosure | High | Medium | Query rate limiting; canary inference queries | Platform |
| ML-T004 | ML credentials (Hugging Face, OpenAI) | AML.T0012 Valid Accounts | Spoofing | Critical | High | Honeypot API keys + short-lived credentials | SecOps |
| ML-T005 | Training pipeline (Kubeflow) | AML.T0010.001 AI Supply Chain Compromise: AI Software | Tampering | High | Low | Dependency pinning + SBOM | DevSecOps |
| ML-T006 | Feature store (Redis) | AML.T0043 Craft Adversarial Data | Tampering | Medium | Low | Feature value range validation + anomaly alert | Data |

## Data Flow Trust Boundary Analysis

```
[External Data Feeds] ──UNTRUSTED──▶ [Ingestion Layer] ──VALIDATION──▶ [Feature Store]
                                                                                │
                                                                          TRUSTED
                                                                                │
[ML Credentials]  ─────────────────────────────────────────────────▶  [Training Job]
(Vertex AI / SageMaker / openssl-wrapped API keys)                            │
                                                                          ARTIFACT
                                                                                │
[Model Registry] ◀─────────────────────────────────────────────────────────────
(Hugging Face Hub / MLflow / S3)
       │
  SEMI-TRUSTED
       │
[Inference API] ──nginx GATEWAY──▶ [External Consumers]  ←── ADVERSARY THREAT SURFACE
```

## Adversary Objective Mapping

| Adversary Goal | Attack Path | ATLAS Techniques Involved |
|---------------|-------------|--------------------------|
| Steal model IP | Access model registry → copy weights | AML.T0012 Valid Accounts → AML.T0035 AI Artifact Collection → AML.T0025 Exfiltration via Cyber Means |
| Degrade model accuracy | Poison training data in S3 | AML.T0002 Acquire Public AI Artifacts → AML.T0010.002 AI Supply Chain Compromise: Data → AML.T0020 Training Data Poisoning |
| Evade detection | Craft adversarial inputs to bypass security model | AML.T0043 Craft Adversarial Data → AML.T0015 Evade AI Model |
| Reconstruct training data | Membership inference via inference API | AML.T0040 AI Model Inference API Access → AML.T0024.000 Infer Training Data Membership |
| Backdoor model | Insert trigger pattern during fine-tuning on Hugging Face | AML.T0018 Manipulate AI Model → AML.T0015 Evade AI Model |

## ATT&CK Enterprise Mappings for ML Context

| ATT&CK Technique | ML System Context | Detection |
|-----------------|-------------------|-----------|
| T1530 Data from Cloud Storage Object | S3 model weights exfiltration | CloudTrail GetObject on model bucket |
| T1552.005 Cloud Instance Metadata | EC2/GCE metadata service for training credential theft | IMDS access log + restrict to IMDSv2 |
| T1078.004 Cloud Accounts | Compromised SageMaker / Vertex AI / Azure ML service accounts | IAM anomaly detection |
| T1195 Supply Chain Compromise | Poisoned PyPI package injected into training pipeline | SBOM diff + dep pinning |
| T1048 Exfiltration Over Alternative Protocol | Exfiltrate model weights via DNS or openssl-encrypted channel | DNS anomaly + data volume baseline |
| T1119 Automated Collection | Automated inference API queries for model extraction | API rate anomaly detection |

---

## Phase: ml-defense-controls

# ML Defense Controls — Reference

Use during Phase 4 to select and implement technical controls that defend the AI/ML
system against ATLAS-mapped threats identified in Phase 3.

## Defense Control Categories

### 1. Training Data Integrity

| Control | Implementation | Threat Mitigated | Detection Mechanism |
|---------|---------------|-----------------|---------------------|
| Canary samples | Inject ≥50 uniquely-identifiable data points; hash all training samples | AML.T0020 Training Data Poisoning | Monitor model output on canary inputs post-training |
| Data provenance tracking | Hash all training data files (SHA-256); store in apache-style access log | AML.T0115.000 Publish Poisoned AI Artifacts: Datasets | File integrity monitoring on training S3 bucket |
| Differential privacy | Apply ε-DP (ε ≤ 1.0) via TensorFlow Privacy or PyTorch Opacus | AML.T0024.001 Invert AI Model | Statistical audit of model outputs |
| SBOM for training deps | Generate CycloneDX SBOM for all Python/conda packages on linux training host | AML.T0010.001 AI Supply Chain Compromise: AI Software | Dep diff alert on CI pipeline |
| Training data checksums | Verify SHA-256 checksums of all training batches before training begins | AML.T0020 Training Data Poisoning | Hash mismatch alert in training orchestrator |

### 2. Model Registry Protection

| Control | Implementation | Threat Mitigated |
|---------|---------------|-----------------|
| Signed model artifacts | GPG/openssl sign all model weights before publishing to registry | AML.T0018 Manipulate AI Model (backdoor ML model) |
| Immutable registry | Versioned, write-once model storage (S3 Object Lock / MLflow artifact locking) | AML.T0018 Manipulate AI Model |
| Registry access logging | CloudTrail / audit log on all model.download() calls in Hugging Face or MLflow | AML.T0024.002 Extract AI Model |
| Honeypot model version | Publish deliberately degraded model as a canary — any access to this version fires | AML.T0024.002 Extract AI Model |
| Model fingerprinting | Embed unique watermark in model outputs to detect unauthorized copies | AML.T0025 Exfiltration via Cyber Means (exfiltrate ML model) |

### 3. Inference API Hardening

| Control | Tool / Implementation | Threat Mitigated | ATT&CK Analogue |
|---------|----------------------|-----------------|----------------|
| Rate limiting | nginx rate_limit_zone + burst control; or API gateway with per-consumer quotas | AML.T0024.002 Extract AI Model | T1499 |
| Query logging with anomaly detection | Log all inference requests; baseline per-client query volumes; alert on deviations | AML.T0024.002 Extract AI Model, AML.T0040 AI Model Inference API Access | T1119 |
| Output perturbation | Add calibrated noise to high-confidence outputs (reduces inversion accuracy by ≥70%) | AML.T0024.001 Invert AI Model | T1040 |
| canary inference queries | Register specific input strings; fire alert if received at inference endpoint | AML.T0024.002 Extract AI Model, AML.T0015 Evade AI Model | T1102 |
| TLS certificate pinning | Client certificate required for high-privilege inference; openssl client auth | AML.T0040 AI Model Inference API Access | T1599 |
| Input validation | Schema validation + perturbation detection (L-inf norm check) for adversarial inputs | AML.T0043 Craft Adversarial Data, AML.T0015 Evade AI Model | T1203 |

### 4. ML Credential Security

| Control | Implementation | Threat Mitigated | ATT&CK Analogue |
|---------|---------------|-----------------|----------------|
| Honeytoken API keys | Create fake API keys for Vertex AI, SageMaker, Hugging Face, OpenAI; alert on any use | AML.T0012 Valid Accounts | T1078 |
| Short-lived credentials | Rotate ML platform API keys every 24h; use OIDC federation where possible | AML.T0012 Valid Accounts | T1552 |
| Vault for ML secrets | Store all Vertex AI, SageMaker, Hugging Face tokens in HashiCorp Vault on linux | AML.T0012 Valid Accounts | T1552.001 |
| Credential scanning | Run detect-secrets / trufflehog on all Git repos; block commits containing API keys | AML.T0012 Valid Accounts | T1195 |
| Least-privilege IAM | ML training role: read training bucket only; inference role: invoke model endpoint only | AML.T0012 Valid Accounts | T1548 |

### 5. Supply Chain Controls

| Control | Implementation | ATLAS Threat | ATT&CK Analogue |
|---------|---------------|-------------|----------------|
| Pinned dependencies | requirements.txt with exact hashes (`pip install --require-hashes`) | AML.T0010.001 AI Supply Chain Compromise: AI Software | T1195 |
| Private PyPI mirror | Internal apache-served PyPI mirror; allowlist of approved packages | AML.T0010.001 AI Supply Chain Compromise: AI Software | T1195.001 |
| Pre-trained model verification | Verify SHA-256 hash of downloaded Hugging Face / PyTorch model before use | AML.T0002 Acquire Public AI Artifacts | T1588 |
| CI pipeline isolation | Training jobs run in ephemeral linux containers; no internet egress during training | AML.T0015 Evade AI Model | T1562 |
| SBOM generation + diff | CycloneDX SBOM on every training image build; alert on unexpected dep additions | AML.T0010.001 AI Supply Chain Compromise: AI Software | T1195 |

## Control Implementation Priority

```
Priority 1 — Immediate (week 1):
├── Honeytoken ML credentials for all ML platforms
├── nginx rate limiting on inference API
├── Canary samples in training data (inject before next training run)
└── Registry access logging (CloudTrail / MLflow audit log)

Priority 2 — Short-term (month 1):
├── Differential privacy (ε-DP) for sensitive model training
├── Signed model artifacts (openssl GPG signing)
├── Canary inference query registration
└── Dependency pinning + private PyPI mirror

Priority 3 — Medium-term (quarter 1):
├── Output perturbation for high-risk inference APIs
├── SBOM generation in CI/CD pipeline
└── Input adversarial perturbation detection
```

## Security Metrics for ML Controls

| Metric | Target | Measurement Frequency |
|--------|--------|-----------------------|
| Honeytoken API key trigger rate | 0 (any trigger = incident) | Real-time |
| Canary sample survival rate in training | 100% (canaries must survive to training) | Per training run |
| Inference API query rate anomalies | < 5 alerts / week (tuned baseline) | Daily |
| Model registry access by non-approved principals | 0 | Daily |
| Dep pin compliance | 100% of training Docker images | Per build |
| Time-to-detect (TTD) for model extraction attempt | < 30 minutes | Per incident |

---

## Phase: ml-incident-response

# ML Incident Response — Reference

Use during Phase 5 when an ML security incident has been detected. Covers containment,
investigation, and recovery for the most common ATLAS attack scenarios.

## ML Incident Classification

| Incident Type | Detection Signal | ATLAS Technique | Severity |
|---------------|-----------------|-----------------|----------|
| Training data poisoning confirmed | Canary sample appears in model output; accuracy degradation on clean test set | AML.T0020 | Critical |
| Model extraction in progress | API rate anomaly; canary inference query triggered | AML.T0024.002 Extract AI Model | High |
| Model inversion attack | Repeated queries from same source; training data identified in API responses | AML.T0024.001 Invert AI Model | High |
| ML credential compromise | Honeytoken API key triggered (Vertex AI, SageMaker, Hugging Face, OpenAI) | AML.T0012 Valid Accounts | Critical |
| Model backdoor discovered | Model performance differs on triggered vs. non-triggered inputs | AML.T0018 Manipulate AI Model | Critical |
| Adversarial example campaign | Production model confidence drops; evasion rate increases | AML.T0015 Evade AI Model | Medium |
| Supply chain compromise | SBOM diff shows unexpected dependency; training pipeline compromised on linux host | AML.T0010.001 AI Supply Chain Compromise: AI Software | Critical |

## Incident Response Playbooks

### Playbook 1: Training Data Poisoning (AML.T0020)

**Trigger:** Canary sample appears in model outputs, or accuracy unexpectedly drops on clean validation set.

```
CONTAIN:
1. Halt all pending training runs immediately
2. Quarantine training data source (suspend S3 bucket write access; freeze apache data pipeline)
3. Snapshot current model weights for forensic analysis

INVESTIGATE:
4. Compare training data checksums against known-good baseline
5. Identify timestamp range of injected samples — query S3 access logs for write operations
6. Run canary sample detection: `python canary_check.py --model ./model.pkl --canaries canaries.json`
7. Check SBOM diff for malicious package added to training pipeline

RECOVER:
8. Roll back training data to last known-good snapshot (S3 versioning)
9. Re-train model from clean dataset with verified hashes
10. Run full canary sample suite on retrained model before promoting to production
11. Update incident report with T1565.001 (Data Manipulation) ATT&CK mapping
```

### Playbook 2: ML Credential Compromise (AML.T0012 Valid Accounts)

**Trigger:** Honeytoken API key used (Vertex AI, SageMaker, Hugging Face, OpenAI), or legitimate API key appears in SIEM from unexpected IP.

```
CONTAIN (within 15 minutes):
1. Revoke compromised API key immediately across all ML platforms:
   - Vertex AI: gcloud iam service-accounts keys delete <KEY_ID>
   - SageMaker: aws iam delete-access-key --access-key-id <KEY_ID>
   - Hugging Face: Organization settings → Access Tokens → revoke the token
   - OpenAI: Platform → API Keys → Revoke

2. Rotate ALL sibling credentials (same service account family)
3. Block source IP at nginx / API gateway level

INVESTIGATE:
4. Enumerate all API calls made with compromised credential — CloudTrail / audit log
5. Check if model weights were downloaded (S3 GetObject on model bucket)
6. Check if any training jobs were started (check for AML.T0020 Training Data Poisoning follow-on)
7. Determine exfiltration path (T1537 Transfer to Cloud, T1041 Exfiltration over C2)
8. Identify initial compromise vector — git repo secret exposure, phishing, linux credential theft

RECOVER:
9. Issue new credentials with least-privilege scoping
10. Store in HashiCorp Vault; enable dynamic secret rotation
11. Deploy new honeytoken for same platform to maintain detection coverage
```

### Playbook 3: Model Extraction (AML.T0024.002 Extract AI Model)

**Trigger:** API rate anomaly; canary inference query triggered; unusually consistent query patterns from one source.

```
CONTAIN:
1. Block source IP/ASN at nginx WAF layer
2. Enable aggressive rate limiting: 10 req/min per IP
3. Enable output perturbation: add ε=0.01 noise to all confidence scores

INVESTIGATE:
4. Extract query log for source IP — how many queries, what input distribution?
5. Estimate extraction progress: N_queries × model_size_proxy → coverage estimate
6. Check if queries target specific decision boundaries (sign of active learning attack)
7. ATT&CK mapping: T1119 Automated Collection → T1048 Exfiltration

RECOVER:
8. Embed model watermark in newly promoted model version
9. File IP block at perimeter (cisco ASA / fortinet fortigate / paloaltonetworks PAN-OS)
10. If extraction near-complete: consider architectural changes (larger output perturbation, ensemble outputs)
```

### Playbook 4: Model Backdoor (AML.T0018 Manipulate AI Model)

**Trigger:** Model behaves differently when trigger pattern present; abnormal behaviour on specific input categories.

```
CONTAIN:
1. Immediately roll back inference API to previous model version
2. Quarantine affected model version in registry (mark as COMPROMISED)
3. Alert all consumers of the model — cease using predictions

INVESTIGATE:
4. Compare model weights between compromised and known-good version (diff sha256 hashes)
5. Run trigger-pattern analysis: test known adversarial trigger patterns against model
6. Inspect training pipeline git history for unauthorised changes — check linux runner logs
7. Inspect model registry for suspicious uploads from non-approved principals

RECOVER:
8. Re-train from earliest known-clean training data snapshot
9. Implement model signing (GPG / openssl signature) before registry push
10. Add training pipeline integrity check as CI gate
```

## Forensic Evidence Preservation

For any ML incident, preserve the following before remediation:

| Artefact | Location | Preservation Method |
|----------|----------|---------------------|
| Model weights (compromised) | MLflow / S3 model bucket | S3 Object Lock copy to forensic bucket |
| Training data at time of incident | S3 training bucket | S3 versioned snapshot |
| API query logs (nginx access.log) | /var/log/nginx/access.log | Copy to SIEM + S3 |
| CloudTrail logs (S3, SageMaker, Vertex AI) | CloudTrail S3 bucket | Verify log integrity hashes |
| linux training host memory dump | /proc/PID/mem of training process | Volatile memory capture (LiME kernel module) |
| Container image used for training | Docker registry | docker save + sha256 hash |
| Git commit history of training code | GitHub / GitLab | git log --all with GPG signature check |

## ATT&CK Technique Coverage — ML Incident Response

T1078 T1078.004 T1119 T1195 T1195.001 T1530 T1537 T1540 T1548 T1552 T1552.005
T1562 T1565 T1565.001 T1588 T1588.002 T1048 T1041 T1499 T1602 T1213
T1040 T1102 T1059 T1059.006 T1203 T1600 T1119 T1074 T1574 T1574.001

---

## Phase: live-threat-intel

# Live Threat Intel — mitre-atlas

<!-- BEGIN AEGIS-INTEL-SYNC — auto-generated, do not edit by hand -->

_Auto-generated by `scripts/intel-sync.js` on 2026-08-22 from the operator's threat-intel corpus (window: last 7 days). Regenerable — do not hand-edit inside these markers. Human-authored tradecraft lives in the other reference files and is never modified by this process._

## Current Observations (2026-08-22)

**Routed to this skill:** 41 feed item(s), 0 teardown(s).

| Date | Observation | CVEs | Actors | Source |
|------|-------------|------|--------|--------|
| 2026-08-13 | From Prompt Injection to Web Exploitation: A Component Taxonomy for LLM2X Attacks | — | — | arXiv |
| 2026-08-08 | PIMiner: An Agentic System for Automatic Prompt-Injection Red Teaming | — | — | arXiv |
| 2026-08-06 | AgentAntibody: A Self-Evolving Immune System for Defending LLM Agents Against Prompt Injection | — | — | arXiv |
| 2026-08-16 | CyberStag-Security-26B: New Open-Weight Cybersecurity Reasoning Model on Hugging Face | — | — | HuggingFace |
| 2026-08-16 | THN Weekly Recap: AI Goes Rogue, Metabase 0-Day, MCP Supply-Chain Attacks, and Router Backdoors | — | — | The Hacker News |
| 2026-08-11 | Show HN: Needle2, a 14MB Agentic LLM for Phones, Wearables, and Robots | — | — | Hacker News |
| 2026-08-11 | Palo Alto Researcher Demos ChatGPT Sandbox C2 Chain at Black Hat | — | — | Dark Reading |
| 2026-07-30 | Anthropic Discloses Three AI Agent Sandbox Escapes Tied to Evaluation Vendor Irregular | — | — | Cyber Unit |
| 2026-08-05 | Meta's Muse Spark 1.1 Model Escaped Its Sandbox and Breached an Unnamed Company | — | — | Dark Reading |
| 2026-07-23 | AgentForger Flaw Let a Single Phishing Link Spin Up a Rogue ChatGPT Agent | — | — | Zenity Labs / The Hacker News |
| 2026-08-10 | Red Team Pushes Fake AI Agent Skill to ~26,000 Agents via Marketplace and Instagram Ad | — | — | Rock Cyber Musings |
| 2026-08-14 | Cisco Releases Antares: Open-Weight SLMs for Vulnerability Localization | — | — | Cisco Blogs |

### CVEs referenced

CVE-2026-25536, CVE-2026-33032, CVE-2025-32711, CVE-2026-50548, CVE-2026-50549, CVE-2026-21520, CVE-2026-13242, CVE-2026-55803

### Trending in the knowledge graph (relevant to this surface)

Prompt Injection (seen 114x, last 2026-08-13) · Jailbreak (seen 58x, last 2026-08-13) · Microsoft (seen 207x, last 2026-08-13) · Google (seen 149x, last 2026-08-13) · Anthropic (seen 77x, last 2026-08-13) · Openai (seen 68x, last 2026-08-13) · Cisco (seen 54x, last 2026-08-13) · Palo Alto (seen 33x, last 2026-08-12)

### Coverage questions raised by this period

- CVEs referenced this period: CVE-2026-25536, CVE-2026-33032, CVE-2025-32711, CVE-2026-50548, CVE-2026-50549, CVE-2026-21520 — verify whether this estate is exposed
- Dominant themes: paper (x8), prompt-injection (x8), model (x5), sandbox-escape (x5), agent-security (x4), open-weight (x4) — confirm controls address the recurring mechanism, not just individual incidents
- Vendors/products named repeatedly: OpenAI (x5), Anthropic (x5), Irregular (x3), kontext-security (x2) — review exposure to these in the estate

_These are prompts for verification, not confirmed gaps. Confirm against actual controls before treating any as covered or open — an unvalidated detection is partial coverage, not coverage._

<!-- END AEGIS-INTEL-SYNC -->
