Can open-source prompt-injection detectors catch realistic AI agent attacks?

github.com

Can open-source prompt-injection detectors catch realistic AI agent attacks?

github.com

RSS BotBM to Hacker NewsEnglish · 11 days ago

GitHub - rudratoshs/buried-injections: 🛡️ Regex catches 0%, Meta's Prompt Guard 2 catches 1% of 629 realistic AgentDojo injection attacks when they're buried in tool output. Reproducible benchmark.

github.com

🛡️ Regex catches 0%, Meta's Prompt Guard 2 catches 1% of 629 realistic AgentDojo injection attacks when they're buried in tool output. Reproducible benchmark. - rudratoshs/buried-injections

Comments

You must log in or register to comment.

Chat

Hacker News

hackernews

You are not logged in. However you can subscribe from another Fediverse account, for example Lemmy or Mastodon. To do this, paste the following into the search field of your instance: !hackernews@lemmy.bestiver.se

Community locked: only moderators can create posts. You can still comment on posts.

Posts from the RSS Feed of HackerNews.

The feed sometimes contains ads and posts that have been removed by the mod team at HN.

Source of the RSS Bot

Visibility: Public

This community can be federated to other instances and be posted/commented in by their users.

528 users / day
1.65K users / week
2.53K users / month
9.05K users / 6 months
2 local subscribers
5.49K subscribers
61.9K Posts
33.3K Comments
Modlog