FeatureLens: A Highly Generalizable and Interpretable Framework for Detecting Adversarial Examples Based on Image Features

2025-12-03

AI Summary

Deep neural network (DNN) image classifiers are vulnerable to adversarial attacks, while existing detection methods suffer from poor interpretability and limited generalization. To address this, we propose FeatureLens—a lightweight, highly interpretable adversarial sample detection framework. FeatureLens employs a shallow feature extractor (51-dimensional) coupled with lightweight classifiers (SVM, MLP, or XGBoost), containing only 1K–30K parameters, thereby enhancing both interpretability and deployment efficiency. Under diverse attacks—including FGSM, PGD, CW, and DAmageNet—it achieves closed-set detection accuracy of 97.8%–99.75% and cross-attack generalization performance of 86.17%–99.6%. The framework is robust, computationally efficient, and model-agnostic. Its core innovation lies in leveraging ultra-low-dimensional semantic features for detection, breaking reliance on high-dimensional “black-box” representations and establishing a novel paradigm for trustworthy AI defense.

Technology Category

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

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Although the remarkable performance of deep neural networks (DNNs) in image classification, their vulnerability to adversarial attacks remains a critical challenge. Most existing detection methods rely on complex and poorly interpretable architectures, which compromise interpretability and generalization. To address this, we propose FeatureLens, a lightweight framework that acts as a lens to scrutinize anomalies in image features. Comprising an Image Feature Extractor (IFE) and shallow classifiers (e.g., SVM, MLP, or XGBoost) with model sizes ranging from 1,000 to 30,000 parameters, FeatureLens achieves high detection accuracy ranging from 97.8% to 99.75% in closed-set evaluation and 86.17% to 99.6% in generalization evaluation across FGSM, PGD, CW, and DAmageNet attacks, using only 51 dimensional features. By combining strong detection performance with excellent generalization, interpretability, and computational efficiency, FeatureLens offers a practical pathway toward transparent and effective adversarial defense.

Problem

• Detects adversarial examples in image classification using interpretable features
• Addresses poor generalization and interpretability in existing detection methods
• Provides lightweight framework for transparent and efficient adversarial defense

Innovation

• Lightweight framework using shallow classifiers for detection
• Extracts only 51-dimensional image features for efficiency
• Achieves high accuracy across multiple adversarial attack types