🤝 Top Collaborators
🤓 Latest Submissions
MAIDEN
It works in 3 layers: first layer: RegEx based heuristics, 2nd layer: Fine tuned Sentence encoder, third layer: fine tuned small language model, run the UI, give the prompt, & know whether it's a MINJA attack or not, simple! It also acts as a basic guardrail, that filters out basic prompt injection attempts. It encounters: Memory alteration, personality alteration, Exfiltration etc. Secure your LLM's memory, one prompt at a time! Tech stack used: Nextjs, python, fastapi, sentence encoder, llm, unsloth etc. Simple usage: Go to UI, type your prompt, and our engine will detect if it is a Minja attack or not!
Hackathon link
19 May 2026
