Self-Healing Robust Neural Networks via Closed-Loop Control
Self-Healing Robust Neural Networks via Closed-Loop Control
复制标题
DOI：
10.48550/arxiv.2206.12963
复制
发表时间：
2022-06
期刊：
影响因子：
--
通讯作者：
Zhuotong Chen;Qianxiao Li;Zheng Zhang
中科院分区：
文献类型：
--
作者：
Zhuotong Chen;Qianxiao Li;Zheng Zhang
神经网络在广泛应用的同时，其脆弱性问题也日益受到关注。虽然已经开发了许多攻击和防御技术，但这项工作从一个新的角度研究了鲁棒性问题：我们能否设计一个自我修复的神经网络，它可以自动检测和修复漏洞问题？典型的自我修复机制是人体的免疫系统。这种受生物学启发的想法已被用于许多工程设计中，但很少在深度学习中进行研究。本文考虑了神经网络的训练后自愈问题，提出了一种闭环控制公式来自动检测和修复各种攻击或扰动引起的误差。我们提供了一个基于边际的分析来解释这个公式如何提高分类器的鲁棒性。为了加快所提出的自愈网络的推理速度，我们通过改进基于庞特里亚金极大原理的求解器来解决控制问题。最后，我们给出了具有非线性激活函数的神经网络框架的误差估计。我们在几种网络架构上针对各种扰动验证了性能。由于自愈方法不需要关于数据扰动/攻击的先验信息，它可以处理广泛的不可预见的扰动。
Despite the wide applications of neural networks, there have been increasing concerns about their vulnerability issue. While numerous attack and defense techniques have been developed, this work investigates the robustness issue from a new angle: can we design a self-healing neural network that can automatically detect and fix the vulnerability issue by itself? A typical self-healing mechanism is the immune system of a human body. This biology-inspired idea has been used in many engineering designs but is rarely investigated in deep learning. This paper considers the post-training self-healing of a neural network, and proposes a closed-loop control formulation to automatically detect and fix the errors caused by various attacks or perturbations. We provide a margin-based analysis to explain how this formulation can improve the robustness of a classifier. To speed up the inference of the proposed self-healing network, we solve the control problem via improving the Pontryagin Maximum Principle-based solver. Lastly, we present an error estimation of the proposed framework for neural networks with nonlinear activation functions. We validate the performance on several network architectures against various perturbations. Since the self-healing method does not need a-priori information about data perturbations/attacks, it can handle a broad class of unforeseen perturbations.