AEGIS: A Semantic GAN and Evidential Learning Frameworkfor Robust Adversarial Detection in Vision Sensors

2026-06-25

AI Summary

This work addresses the vulnerability of deep neural networks to imperceptible adversarial perturbations that induce erroneous predictions in visual tasks. The authors propose a semantic-aware and uncertainty-guided framework for adversarial example detection, which leverages SemantiGAN to filter out semantically inconsistent inputs. Robust classification and calibrated uncertainty estimation are achieved by integrating stochastic augmentations, multi-view prediction inconsistency—quantified via FlipScore, inter-layer cosine similarity, and entropy—and evidential deep learning based on Dirichlet distributions. Evaluated on Tiny ImageNet against six attack types, the method attains an AUROC of 92.1%, AUPRC of 90.2%, and classification accuracy of 90.7%, substantially outperforming conventional softmax-based baselines while offering strong detection performance, interpretability, and reliable uncertainty quantification.

Abstract

Deep neural networks (DNNs) have shown outstanding performance in visual recognition tasks within vision sensor networks; however, they are still vulnerable to adversarial manipulations and imperceptible perturbations that can lead to erroneous predictions. To address that, this paper presents AEGIS, a semantic aware and uncertainty guided adversarial detection framework designed for robust image classification in vision sensors pipelines. At its core, a SemantiGAN module functions as a multi class semantic discriminator, identifying and filtering visually inconsistent adversarial inputs before they propagate further in the pipeline. For inputs that pass this stage, a stochastic augmentation process generates test time variations, from which handcrafted instability metrics FlipScore, Prediction Inconsistency, Layerwise Cosine Similarity (early and mid layers), and Entropy are computed. These features are aggregated into a compact five dimensional vector and processed by an Evidential Deep Learning (EDL) classifier, which models output evidence using a Dirichlet distribution to yield both class predictions and calibrated uncertainty estimates. Evaluations on the Tiny ImageNet dataset across six categories clean, FGSM, PGD, patch based, functional, and geometric attacks demonstrate the effectiveness of AEGIS. The proposed framework achieves an AUROC of 92.1\%, an AUPRC of 90.2\%, and an accuracy of 90.7\%, outperforming conventional softmax-based detectors in terms of detection performance, robustness, interpretability, and uncertainty calibration.

Problem

• adversarial detection
• vision sensors
• deep neural networks
• semantic inconsistency
• uncertainty calibration

Innovation

• Semantic GAN
• Evidential Deep Learning
• Adversarial Detection
• Uncertainty Calibration
• Test-Time Augmentation