Drive-by Compromise
Description
Adversaries may gain access to an AI system through a user visiting a website over the normal course of browsing, or an AI agent retrieving information from the web on behalf of a user. Websites can contain an LLM Prompt Injection which, when executed, can change the behavior of the AI model.
The same approach may be used to deliver other types of malicious code that don't target AI directly (See Drive-by Compromise in ATT&CK).
- ATLAS ID
- AML.T0078
- Maturity
- DemonstratedShown to work in research or red-team exercises.
- Platforms
- Enterprise
Mitigations
- AML.M0020Generative AI Guardrails
Apply retrieval, input, and tool-action guardrails to detect and block malicious instructions embedded in websites before execution.
Case studies that use it
Related ATT&CK technique
ATLAS adapts this technique from MITRE ATT&CK Enterprise.
Across MITRE's defensive frameworks
Linked through the 1 ATT&CK technique above.
Counter with D3FEND
Defensive techniques MITRE D3FEND maps to the same ATT&CK techniques.
- D3-IAAIdentifier Activity Analysis1 shared
- D3-UAURL Analysis1 shared
- D3-URAURL Reputation Analysis1 shared
- D3-HDHomoglyph Detection1 shared
- D3-PSEPProcess Segment Execution Prevention1 shared
- D3-SAORSegment Address Offset Randomization1 shared
- D3-APCAApplication Protocol Command Analysis1 shared
- D3-CSPPClient-server Payload Profiling1 shared
- D3-NTCDNetwork Traffic Community Deviation1 shared
- D3-NTFNetwork Traffic Filtering1 shared
- D3-NTSANetwork Traffic Signature Analysis1 shared
- D3-PHDURAPer Host Download-Upload Ratio Analysis1 shared