SIGNAL//DESK
AI securitysrc: MITRE ATLAS

Impersonation

Impersonation is when a scammer pretends to be someone you trust—like a boss, a coworker, or a company you use—to trick you into doing something for them, such as sharing a password or sending money. It is like a stranger wearing a mask of your friend to get you to open the door.

Impersonation is a social engineering tactic where an adversary assumes the identity of a trusted entity, such as an executive or vendor, to manipulate a target into performing unauthorized actions. In AI security, this often involves using LLMs to craft highly convincing phishing communications to gain access to sensitive resources like model repositories or container registries.

Impersonation is a deceptive technique wherein an adversary adopts the persona of a trusted individual or organization to facilitate social engineering, specifically to subvert human trust for unauthorized access or privilege escalation. Within the AI security lifecycle, this includes the use of generative models to automate spearphishing campaigns that target personnel with access to AI DevOps infrastructure, such as model registries, container registries, and software supply chain components, to achieve persistent access or exfiltration.


← all terms