Feature Compression May Be the Root Cause of Adversarial Fragility in Neural Network Classifiers (Student Abstract)

Jingchao Gao; Ziqing Lu; Raghu Mudumbai; Xiaodong Wu; Jirong Yi; Myung Cho; Catherine Xu; Hui Xie; Weiyu Xu

doi:10.1609/aaai.v40i48.42217

Back

Feature Compression May Be the Root Cause of Adversarial Fragility in Neural Network Classifiers (Student Abstract)

Journal article

Open access

Feature Compression May Be the Root Cause of Adversarial Fragility in Neural Network Classifiers (Student Abstract)

Jingchao Gao, Ziqing Lu, Raghu Mudumbai, Xiaodong Wu, Jirong Yi, Myung Cho, Catherine Xu, Hui Xie and Weiyu Xu

Proceedings of the ... AAAI Conference on Artificial Intelligence, Vol.40(48), pp.41212-41213

03/14/2026

DOI: 10.1609/aaai.v40i48.42217

Files and links (1)

url

https://doi.org/10.1609/aaai.v40i48.42217View

Published (Version of record) Open Access

Abstract

In this paper, we study the adversarial robustness of deep neural networks (DNN) for classification against optimal classifiers. We look at the smallest magnitude of possible additive perturbations that can change a classifier's output. We provide a matrix-theoretic explanation of the adversarial fragility of DNNs for classification. In particular, our theoretical results show that the adversarial robustness of a neural network can degrade as the input dimension d increases. Analytically, we show that the adversarial robustness of neural networks can be only 1/√d of the best possible adversarial robustness of optimal classifiers. Our theories match remarkably well with empirical results. The matrix-theoretic explanation aligns with an earlier information-theoretic feature-compression-based explanation for the adversarial fragility of neural networks.

Details

Title: Subtitle: Feature Compression May Be the Root Cause of Adversarial Fragility in Neural Network Classifiers (Student Abstract)
Creators: Jingchao Gao - Minnesota State Colleges and Universities System
Ziqing Lu - University of Iowa
Raghu Mudumbai - University of Iowa
Xiaodong Wu - University of Iowa
Jirong Yi - University of Iowa
Myung Cho - University of Iowa
Catherine Xu
Hui Xie - University of Iowa
Weiyu Xu - University of Iowa
Resource Type: Journal article
Publication Details: Proceedings of the ... AAAI Conference on Artificial Intelligence, Vol.40(48), pp.41212-41213
DOI: 10.1609/aaai.v40i48.42217
ISSN: 2159-5399
eISSN: 2374-3468
Number of pages: 2
Language: English
Date published: 03/14/2026
Academic Unit: Electrical and Computer Engineering; Radiation Oncology; The Iowa Institute for Biomedical Imaging
Record Identifier: 9985149708102771

Metrics

1 Record Views