Desk/2024-W06

Week of Feb 5–11, 2024

1 records0 status changes on new evidence1 new findings

New findings

Feb 9, 2024

Published prompt-injection defenses report attack success cut to near zero, or under 10%, against most of the fixed attacks their authors tested.

Revalidatemeasured4 evidence records1 linked finding

Defense & research

Feb 9, 2024

StruQ proposes separating prompts and data channels to defend against prompt injection

DefensePaperUC Berkeley

Chen, Piet, Sitawarin and Wagner propose structured queries, in which a secure front-end separates the trusted prompt from untrusted data and a model fine-tuned to ignore instructions appearing in the data portion. The paper reports much better resistance to prompt injection with little utility loss; it appeared at USENIX Security 2025.