Detecting hidden attacks in AI models by tracking internal changes

Original: Detecting Trojaned DNNs via Spectral Regression Analysis

Source: arxiv.org ↗