A Low-cost Strategic Monitoring Approach for Scalable and Interpretable Error Detection in Deep Neural Networks

Florian Geissler, Syed Qutub, Michael Paulitsch and Karthik Pattabiraman, To appear in the proceedings of the International Conference on Computer Safety, Reliability and Security (SafeComp), 2023. (Acceptance Rate: TBD) [ PDF | Talk ]

Abstract: We present a highly compact run-time monitoring approach for deep computer vision networks that extracts selected information from only a few (down to merely two) hidden layers, yet can efficiently detect silent data corruption originating from both hardware memory and input faults. Building on the insight that critical faults typically manifest as peak or bulk shifts in the activation distribution of the affected network layers, we use strategically placed quantile markers to make accurate estimates about the anomaly of the current inference as a whole. Our method illustrates that knowledge contained in intermediate layer activations can be compressed by three to six orders of magnitude without losing characteristic signatures hinting at underlying system faults. Importantly, the detector component itself is kept algorithmically transparent to render the categorization of regular and abnormal behavior interpretable to a human. Our technique achieves up to 96% precision and 98% recall of detection. Compared to state-of-the-art anomaly detection techniques, this approach requires minimal compute overhead (as little as 0.3% with respect to non-supervised inference time) and contributes to the explainability of the model.

Comments are closed.