{
  "id": "llmjacking-sysdig-2024-05",
  "name": "LLMjacking via stolen cloud credentials (Sysdig)",
  "date_reported": "2024-05-06",
  "researcher": "Alessandro Brucato / Sysdig TRT",
  "org": "Sysdig",
  "sources": [
    {
      "url": "https://www.sysdig.com/blog/llmjacking-stolen-cloud-credentials-used-in-new-ai-attack",
      "title": "Sysdig: LLMjacking - Stolen Cloud Credentials Used in New AI Attack"
    },
    {
      "url": "https://www.sysdig.com/blog/growing-dangers-of-llmjacking",
      "title": "Sysdig: Growing Dangers of LLMjacking"
    },
    {
      "url": "https://www.pillar.io/blog/bizarre-bazaar-llmjacking-jan-2026",
      "title": "Pillar Security: Bizarre Bazaar - LLMjacking Campaign Targeting AI Startups"
    }
  ],
  "magicsword_coverage": [],
  "borderline": false,
  "borderline_note": null,
  "ai_role": "ai_credential_theft",
  "ai_role_secondary": null,
  "ai_role_detail": null,
  "technique_summary": "After a Laravel CVE-2021-3129 foothold, attackers stole cloud credentials and probed ten AI services (AI21, Anthropic, AWS Bedrock, Azure, ElevenLabs, MakerSuite, Mistral, OpenAI, OpenRouter, and GCP Vertex) to resell LLM access via reverse proxies while victims paid inference bills. Later Sysdig reporting showed 85,000+ Bedrock requests with Converse API abuse within 30 days of launch. Pillar Security separately reported the Bizarre Bazaar campaign targeting AI startups for unauthorized inference resale.",
  "initial_access": "Vulnerable Laravel (CVE-2021-3129) leading to cloud credential theft",
  "ai_tooling_and_models": [
    "AWS Bedrock Claude v2/v3",
    "OpenAI",
    "Anthropic",
    "Azure",
    "Vertex AI",
    "others in probe list"
  ],
  "other_tooling": [
    "LLM reverse proxy for resale"
  ],
  "victims_and_sectors": "unknown",
  "regions": [
    "unknown"
  ],
  "attribution_or_cluster": "unknown",
  "scale_and_metrics": "Potential >$46,000/day Claude 2.x costs cited; July 2024 spike 61k Bedrock requests in 3 hours",
  "prevention_angle": "Monitor Bedrock ListFoundationModels+Invoke/Converse from long-term AKIA keys; rotate exposed keys; SCP restrict Bedrock",
  "related_magicsword_projects": [],
  "source_notes": "2026 updates expand LLMjacking scope: Bedrock Marketplace abuse (FortiGuard), AI startup targeting (Pillar Bizarre Bazaar), and session token theft (Okta)",
  "timeline": {
    "first_seen": "2024-05-06",
    "last_updated": "2026-11-05",
    "updates": [
      {
        "date": "2024-05-06",
        "source": "Sysdig",
        "summary": "Initial LLMjacking disclosure; Laravel CVE-2021-3129 to cloud credential theft and AI inference resale"
      },
      {
        "date": "2024-07-01",
        "source": "Sysdig",
        "summary": "Bedrock Converse API abuse observed; 85k+ requests within 30 days of launch"
      },
      {
        "date": "2026-01-28",
        "source": "Pillar Security",
        "summary": "Bizarre Bazaar campaign targets AI startups with LLMjacking"
      },
      {
        "date": "2026-09-03",
        "source": "FortiGuard",
        "summary": "AWS Bedrock Marketplace abuse; attackers subscribe victims to high-cost models without approval"
      },
      {
        "date": "2026-09-09",
        "source": "Okta Security",
        "summary": "Stolen AI session tokens enable LLMjacking without long-term credentials; bypasses MFA and key rotation"
      }
    ]
  },
  "last_reviewed": "2026-11-05",
  "attack_flow": {
    "nodes": [
      {
        "id": "step-1",
        "title": "Vulnerable Laravel provides a foothold",
        "description": "The report identifies CVE-2021-3129 before credential theft.",
        "kind": "access",
        "source_urls": [
          "https://www.sysdig.com/blog/llmjacking-stolen-cloud-credentials-used-in-new-ai-attack",
          "https://www.sysdig.com/blog/growing-dangers-of-llmjacking"
        ]
      },
      {
        "id": "step-2",
        "title": "Cloud credentials are stolen",
        "description": "Attackers obtain keys and probe multiple AI providers.",
        "kind": "execution",
        "source_urls": [
          "https://www.sysdig.com/blog/llmjacking-stolen-cloud-credentials-used-in-new-ai-attack",
          "https://www.sysdig.com/blog/growing-dangers-of-llmjacking"
        ]
      },
      {
        "id": "step-3",
        "title": "Inference access is proxied",
        "description": "Authorized model access is resold through attacker reverse proxies.",
        "kind": "ai",
        "source_urls": [
          "https://www.sysdig.com/blog/llmjacking-stolen-cloud-credentials-used-in-new-ai-attack",
          "https://www.sysdig.com/blog/growing-dangers-of-llmjacking"
        ]
      },
      {
        "id": "step-4",
        "title": "Victim accounts incur inference usage",
        "description": "The report describes high-volume requests and potentially substantial costs.",
        "kind": "impact",
        "source_urls": [
          "https://www.sysdig.com/blog/llmjacking-stolen-cloud-credentials-used-in-new-ai-attack",
          "https://www.sysdig.com/blog/growing-dangers-of-llmjacking"
        ]
      }
    ],
    "edges": [
      {
        "source": "step-1",
        "target": "step-2"
      },
      {
        "source": "step-2",
        "target": "step-3"
      },
      {
        "source": "step-3",
        "target": "step-4"
      }
    ]
  },
  "atlas_techniques": [
    {
      "id": "AML.T0012",
      "rationale": "Stolen cloud credentials are used to access inference services.",
      "source_urls": [
        "https://www.sysdig.com/blog/llmjacking-stolen-cloud-credentials-used-in-new-ai-attack",
        "https://www.sysdig.com/blog/growing-dangers-of-llmjacking"
      ]
    }
  ],
  "category": "attacks-on-ai",
  "ai_involvement_confidence": "confirmed",
  "detections": [
    {
      "name": "Monitor AWS Bedrock API abuse",
      "format": "guidance",
      "notes": "Alert on unusual Bedrock ListFoundationModels, InvokeModel, and Converse API calls, especially from AKIA long-term keys. Monitor for spike patterns (e.g., 61k requests in 3 hours). Implement CloudTrail analysis for Bedrock usage from unexpected geolocations or IAM principals."
    },
    {
      "name": "Detect stolen cloud credentials",
      "format": "guidance",
      "notes": "Scan for exposed AWS keys in public repositories, logs, and error tracking services. Rotate credentials immediately upon detection. Use SCPs to restrict Bedrock access to approved principals only."
    }
  ]
}