QuestionQ24

VerifiedManaging Risks of Generative AI in Software Testing

An attacker deliberately creates prompts that exceed an LLM’s context window, with the aim of overwhelming its memory and unintentionally revealing confidential information from its training data.

Which attack vector type does this scenario BEST describe?

Explanation

Request manipulation involves crafting model inputs to exploit limitations in request handling, such as context-window capacity, in an effort to cause unintended behavior or disclosure.

Community Discussion

No comments yet. Be the first to start the discussion!