Edge-Based Learning for Improved Classification Under Adversarial Noise

2025-04-25

AI Summary

This work addresses the limited robustness of deep learning models against adversarial perturbations, particularly under Fast Gradient Sign Method (FGSM) attacks. We propose a robust learning paradigm grounded in edge features: leveraging Canny edge detection to extract structured, low-level image representations—replacing raw pixels as model inputs—and performing end-to-end training within a ResNet architecture using edge-only inputs. To our knowledge, this is the first systematic demonstration that edge-based representations exhibit significantly higher stability under FGSM compared to pixel-based inputs. Empirical evaluation on brain tumor and COVID-19 medical imaging datasets shows that edge-driven models suffer markedly slower degradation in adversarial accuracy, achieving up to 18.7% improvement in robustness. Our approach reveals the untapped potential of primitive structural features for adversarial defense and establishes a lightweight, module-free, regularization-free robust learning framework.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Adversarial noise introduces small perturbations in images, misleading deep learning models into misclassification and significantly impacting recognition accuracy. In this study, we analyzed the effects of Fast Gradient Sign Method (FGSM) adversarial noise on image classification and investigated whether training on specific image features can improve robustness. We hypothesize that while adversarial noise perturbs various regions of an image, edges may remain relatively stable and provide essential structural information for classification. To test this, we conducted a series of experiments using brain tumor and COVID datasets. Initially, we trained the models on clean images and then introduced subtle adversarial perturbations, which caused deep learning models to significantly misclassify the images. Retraining on a combination of clean and noisy images led to improved performance. To evaluate the robustness of the edge features, we extracted edges from the original/clean images and trained the models exclusively on edge-based representations. When noise was introduced to the images, the edge-based models demonstrated greater resilience to adversarial attacks compared to those trained on the original or clean images. These results suggest that while adversarial noise is able to exploit complex non-edge regions significantly more than edges, the improvement in the accuracy after retraining is marginally more in the original data as compared to the edges. Thus, leveraging edge-based learning can improve the resilience of deep learning models against adversarial perturbations.

Problem

• Studying adversarial noise impact on image classification accuracy
• Exploring edge features' stability under adversarial perturbations
• Evaluating edge-based learning for adversarial attack resilience

Innovation

• Edge-based learning enhances adversarial noise resilience
• Training on edges improves model classification accuracy
• Combining clean and noisy data boosts performance