Beyond Dropout: Robust Convolutional Neural Networks Based on Local Feature Masking

TL;DR

This paper introduces Local Feature Masking (LFM), a novel training strategy for CNNs that improves both robustness against adversarial attacks and generalization by strategically masking features during training.

Contribution

The study proposes LFM, a new method that enhances CNN robustness and generalization by incorporating random feature masking in shallow layers during training.

Findings

LFM significantly improves CNN generalization performance.

LFM enhances CNN resistance to adversarial attacks.

Experimental results show consistent performance gains.

Abstract

In the contemporary of deep learning, where models often grapple with the challenge of simultaneously achieving robustness against adversarial attacks and strong generalization capabilities, this study introduces an innovative Local Feature Masking (LFM) strategy aimed at fortifying the performance of Convolutional Neural Networks (CNNs) on both fronts. During the training phase, we strategically incorporate random feature masking in the shallow layers of CNNs, effectively alleviating overfitting issues, thereby enhancing the model's generalization ability and bolstering its resilience to adversarial attacks. LFM compels the network to adapt by leveraging remaining features to compensate for the absence of certain semantic features, nurturing a more elastic feature learning mechanism. The efficacy of LFM is substantiated through a series of quantitative and qualitative assessments, collectively showcasing a consistent and significant improvement in CNN's generalization ability and resistance against adversarial attacks--a phenomenon not observed in current and prior methodologies. The seamless integration of LFM into established CNN frameworks underscores its potential to advance both generalization and adversarial robustness within the deep learning paradigm. Through comprehensive experiments, including robust person re-identification baseline generalization experiments and adversarial attack experiments, we demonstrate the substantial enhancements offered by LFM in addressing the aforementioned challenges. This contribution represents a noteworthy stride in advancing robust neural network architectures.