Toward Scalable Verification for Safety-Critical Deep Networks
This work addresses safety concerns in applications like autonomous driving and flight control, but it is incremental as it builds on existing verification methods.
The paper tackles the challenge of scaling formal verification for safety-critical deep neural networks, which is currently limited to small systems, by exploring scalable verification techniques and design choices to make systems more verifiable.
The increasing use of deep neural networks for safety-critical applications, such as autonomous driving and flight control, raises concerns about their safety and reliability. Formal verification can address these concerns by guaranteeing that a deep learning system operates as intended, but the state of the art is limited to small systems. In this work-in-progress report we give an overview of our work on mitigating this difficulty, by pursuing two complementary directions: devising scalable verification techniques, and identifying design choices that result in deep learning systems that are more amenable to verification.