Scalable Inference of Symbolic Adversarial Examples
Dimitrov, Dimitar I., Singh, Gagandeep, Gehr, Timon, Vechev, Martin
We present a novel method for generating symbolic adversarial examples: input regions guaranteed to only contain adversarial examples for the given neural network. These regions can generate real-world adversarial examples as they summarize trillions of adversarial examples. We theoretically show that computing optimal symbolic adversarial examples is computationally expensive. We present a method for approximating optimal examples in a scalable manner. Our method first selectively uses adversarial attacks to generate a candidate region and then prunes this region with hyperplanes that fit points obtained via specialized sampling. It iterates until arriving at a symbolic adversarial example for which it can prove, via state-of-the-art convex relaxation techniques, that the region only contains adversarial examples. Our experimental results demonstrate that our method is practically effective: it only needs a few thousand attacks to infer symbolic summaries guaranteed to contain $\approx 10^{258}$ adversarial examples.
Jul-26-2020
- Genre:
- Research Report > New Finding (0.88)
- Industry:
- Government > Military (0.34)
- Information Technology > Security & Privacy (0.48)
- Technology: