一句话：提出 PRODeep 平台，集成约束、抽象与优化三类算法，支持深度神经网络鲁棒性验证与算法组合。
深度神经网络（DNN）已应用于自动驾驶、飞行器避撞、恶意软件检测等安全攸关领域，因此需要保证其鲁棒性，即输入的小扰动不改变输出。然而，已有算法与工具较为分散，难以比较与组合。本文提出 PRODeep，一个面向 DNN 鲁棒性验证的平台。PRODeep 集成了基于约束、基于抽象和基于优化的鲁棒性检查算法，并采用模块化架构，便于对不同算法进行比较。实验部分展示了该工具的使用方式，以及这些技术之间的便捷组合。
Deep neural networks (DNNs) have been applied in safety-critical domains such as self driving cars, aircraft collision avoidance systems, malware detection, etc. In such scenarios, it is important to give a safety guarantee to the robustness property, namely that outputs are invariant under small perturbations on the inputs. For this purpose, several algorithms and tools have been developed recently. In this paper, we present PRODeep, a platform for robustness verification of DNNs. PRODeep incorporates constraint-based, abstraction-based, and optimisation-based robustness checking algorithms. It has a modular architecture, enabling easy comparison of different algorithms. With experimental results, we illustrate the use of the tool, and easy combination of those techniques.
@inproceedings{paperbot5819,
title = {PRODeep: a platform for robustness verification of deep neural networks},
author = {Renjue Li and Jianlin Li and Cheng-Chao Huang and Pengfei Yang and Xiaowei Huang and Lijun Zhang and Bai Xue and Holger Hermanns},
booktitle = {FSE},
year = {2020},
doi = {10.1145/3368089.3417918}
}