Official implementation of paper: DrAttack: Prompt Decomposition and Reconstruction Makes Powerful LLM Jailbreakers
-
Updated
Aug 25, 2024 - JavaScript
Adversarial attacks are techniques that craft intentionally perturbed inputs to mislead machine learning models into producing incorrect outputs. They are central to research in AI robustness, security, and trustworthiness.
Official implementation of paper: DrAttack: Prompt Decomposition and Reconstruction Makes Powerful LLM Jailbreakers
[Tensorflow.js] AdVis: Exploring real-time Adversarial Attacks in the browser with Fast Gradient Sign Method.
A platform that provides users with easy access to AI services developed by Montimage and usage of explainable AI techniques (e.g., LIME, SHAP).
Privacy-first AI Trust & Safety extension that protects AI conversations before they're sent.
Adversarial and Backdoor Attack + Defence
Project Page (PDCL-Attack, ECCV 2024)
Project Page (FACL-Attack, AAAI 2024)
An AI-powered diagnostic tool that classifies brain MRI scans and automatically heals signal interference and noise.
Deliver comprehensive AI-driven security operations with vulnerability scanning, incident response, compliance tracking, and autonomous agents in one platform.