News for hackers. Stuff that matters.

Karan047

all good, but human creativity still hits different
Nothing in cybersecurity is truly solved—only made harder, Prompt injection is no different.

Show HN: Prompt injection detector beats ProtectAI by 19% accuracy, 8.9x smaller

https://huggingface.co/hlyn/prompt-injection-judge-deberta-70m

ProtectAI's deberta-v3 is the most downloaded prompt injection classifier on HuggingFace. We benchmarked directly against it on Qualifire's independent dataset (excluded from our training).* 91.68% vs 72.28% accuracy* 95.84% vs 65.33% precision — they block 1 in 3 legitimate users* 83MB vs 739MB ONNX size* 101ms vs 646ms on CPU

( Read More... | 2 comments | 3 points | on HN )