AgentWatcher

AgentWatcher AgentWatcher is a detection-based defense against indirect prompt injection in LLM agents. It first runs causal context attribution over untrusted context to find the most influential contexts, then applies a monitor LLM that classifies those contexts under explicit, customizable...

CVE-2022-0847-eBPF

CVE-2022-0847-eBPF An eBPF program to detect and defense attacks on CVE-2022-0847. In the kernel space, the program detects splice system calls, if such a system call contains a pipe as the output fd, and the flags of the pipe buffer is set to PIPEBUFFLAGCANMERGE , then an event would be submitte...

detection-defense-library

Detection Defense Library One defense-engineering library, consolidated. Attack classes, the patterns that detect them, the committed Sigma rules, and the Windows trust-boundary model, all in a single repo. This repo folds in five former sibling repositories:...

A Cryptographic Perspective on Mitigation Vs. Detection in Machine Learning

In this paper, we initiate a cryptographically inspired theoretical study of detection versus mitigation of adversarial inputs produced by attackers of Machine Learning algorithms during inference time. We formally define defense by detection DbD and defense by mitigation DbM. Our definitions com...