Results for “regression-detection”

Detect poisoned training data and backdoored models across the ML pipeline using statistical analysis, activation clustering, and spectral signatures.

Identify command-and-control beaconing patterns in network traffic by applying statistical frequency analysis, jitter calculation, and coefficient of variation scoring to detect periodic callbacks from compromised endpoints.