Robust UNet-Based Steganalysis for Secure Image Communication in IoT Camera Systems

Abstract

With the proliferation of Internet of Things (IoT) camera networks, ensuring the security and integrity of image communication has become a critical technical challenge. This paper addresses the problem of robust steganalysis for IoT camera data by presenting an enhanced UNet-based detection framework. The proposed approach integrates adaptive attention modules and adversarial training strategies, enabling precise identification and localization of covert information embedded within digital images. Extensive experiments were conducted on a multi-source dataset incorporating various steganographic techniques and realistic device scenarios. The results demonstrate that the enhanced model achieves a detection accuracy of up to 98% and maintains stable robustness under complex adversarial perturbations, with cross-domain generalization error constrained within 2.7%. Quantitative ablation and comparative studies confirm that architectural innovations in multi-scale feature fusion and adversarial regularization substantially improve both detection reliability and operational applicability in heterogeneous IoT environments. The presented methodology lays a technical foundation for scalable and trustworthy visual forensics solutions, supporting secure and resilient data flows in large-scale, real-world IoT deployments.