SpecRepair: Counter-Example Guided Safety Repair of Deep Neural Networks
It addresses the need for automated safety repair in DNNs for domains like self-driving cars and medical diagnosis, offering an incremental improvement over manual debugging and verification processes.
The paper tackles the problem of automatically repairing unsafe deep neural networks (DNNs) in safety-critical applications by introducing SpecRepair, a tool that efficiently eliminates counter-examples and produces provably safe DNNs without harming classification accuracy, as demonstrated on benchmarks like ACAS Xu and image classification tasks with improved success rates and shorter runtimes compared to existing methods.
Deep neural networks (DNNs) are increasingly applied in safety-critical domains, such as self-driving cars, unmanned aircraft, and medical diagnosis. It is of fundamental importance to certify the safety of these DNNs, i.e. that they comply with a formal safety specification. While safety certification tools exactly answer this question, they are of no help in debugging unsafe DNNs, requiring the developer to iteratively verify and modify the DNN until safety is eventually achieved. Hence, a repair technique needs to be developed that can produce a safe DNN automatically. To address this need, we present SpecRepair, a tool that efficiently eliminates counter-examples from a DNN and produces a provably safe DNN without harming its classification accuracy. SpecRepair combines specification-based counter-example search and resumes training of the DNN, penalizing counter-examples and certifying the resulting DNN. We evaluate SpecRepair's effectiveness on the ACAS Xu benchmark, a DNN-based controller for unmanned aircraft, and two image classification benchmarks. The results show that SpecRepair is more successful in producing safe DNNs than comparable methods, has a shorter runtime, and produces safe DNNs while preserving their classification accuracy.