Formal Verification of CNN-based Perception Systems
This work addresses the verification of neural-based perception systems, which is crucial for safety-critical applications like autonomous vehicles, but it is incremental as it builds on existing reachability analysis methods.
The paper tackles the problem of verifying convolutional neural network (CNN)-based perception systems by defining a new notion of local robustness based on affine and photometric transformations, which cannot be captured by previous robustness notions, and presents a method using reachability analysis and MILP encodings, with experimental results on a CNN trained on the MNIST dataset.
We address the problem of verifying neural-based perception systems implemented by convolutional neural networks. We define a notion of local robustness based on affine and photometric transformations. We show the notion cannot be captured by previously employed notions of robustness. The method proposed is based on reachability analysis for feed-forward neural networks and relies on MILP encodings of both the CNNs and transformations under question. We present an implementation and discuss the experimental results obtained for a CNN trained from the MNIST data set.