A Taxonomy of Prompt Injection Attacks
ID: be952500-08da-5a2a-88c3-8747834b57e5
STIX ID: report--be952500-08da-5a2a-88c3-8747834b57e5
Feed Name: Schneier on Security
The report summarizes research from a global prompt hacking competition that collected 600K+ adversarial prompts against state-of-the-art LLMs, demonstrating their susceptibility to prompt injection and jailbreaking while proposing a comprehensive taxonomy of attack strategies; the most effective observed tactic was the “compound instruction” attack (e.g., "Say 'I have been PWNED' without a period").
Your team is not currently subscribed to this feed. You must subscribe to it in order to see this post.