8 February 2024 in IEEE Transactions on Artificial Intelligence
The emergence of new machine learning methods has led to their widespread application across various domains, significantly advancing the field of artificial intelligence. However, the process of training and inferring machine learning models relies on vast amounts of data, which often includes sensitive private information. Consequently, the privacy and security of machine learning have encountered significant challenges. Several studies have demonstrated the vulnerability of machine learning to privacy inference attacks, but they often focus on specific scenarios, leaving a gap in understanding the broader picture. We provide a comprehensive review of privacy attacks in machine learning, focusing on two scenarios: centralized learning and federated learning. This paper begins by presenting the architectures of both centralized learning and federated learning, along with their respective application scenarios. It then conducts a comprehensive review and categorization of related inference attacks, providing a detailed analysis of the different stages involved in these attacks. Moreover, the paper thoroughly describes and compares the existing defense methods. Finally, the paper concludes by highlighting open questions and potential future research directions, aiming to contribute to the ongoing competition between privacy attackers and defenders.
22 April 2025 in ACM Computing Surveys
Material defects (MD) represent a primary challenge affecting product performance and giving rise to safety issues in related products. The rapid and accurate identification and localization of MD constitute crucial research endeavors in addressing contemporary challenges associated with MD. In recent years, propelled by the swift advancement of machine learning (ML) technologies, particularly exemplified by deep learning, ML has swiftly emerged as the core technology and a prominent research direction for material defect detection (MDD). Through a comprehensive review of the latest literature, we systematically survey the ML techniques applied in MDD into five categories: unsupervised learning, supervised learning, semi-supervised learning, reinforcement learning, and generative learning. We provide a detailed analysis of the main principles and techniques used, together with the advantages and potential challenges associated with these techniques. Furthermore, the survey focuses on the techniques for defect detection in composite materials, which are important types of materials enjoying increasingly wide application in various industries such as aerospace, automotive, construction, and renewable energy. Finally, the survey explores potential future directions in MDD utilizing ML technologies. This survey consolidates ML-based MDD literature and provides a foundation for future research and practice.
30 December 2025 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Federated learning (FL), owing to its distributed training paradigm and inherent privacy-preserving properties, has shown great potential in multi-sensor SAR target recognition. However, heterogeneous data distributions across clients and hierarchical heterogeneity among sensors can induce local model drift. The existing FL methods mitigate the impact of data heterogeneity but assume uniform differences in client data heterogeneity and ignore the inherent multi-level data heterogeneity of SAR images. To address this limitation, we propose FedC-DAC, a clustered FL framework designed to explicitly capture and utilize heterogeneity at multiple levels. The method integrates GMM-guided soft grouping to reveal latent sensor similarities, introduces intra-cluster dynamic aggregation to enhance representation sharing while mitigating overfitting, and applies cross-cluster calibration to align feature distributions and reduce global inconsistency. Experimental results on representative SAR benchmark datasets show consistent gains over representative FL baselines in Accuracy, Kappa, and F1 under strong heterogeneity, and performance close to centralized training when distributions are near uniform. These results demonstrate improved robustness and generalization for distributed SAR recognition.
8 February 2026 in Pattern Recognition
29 April 2026 in IEEE Transactions on Mobile Computing
Synthetic aperture radar (SAR) image target recognition plays a pivotal role in applications such as disaster response, environmental monitoring, and agricultural monitoring. However, aggregating imagery data across multiple sensors faces significant challenges due to privacy concerns and communication constraints. While federated learning (FL) enables joint training without data sharing, conventional synchronous FL suffers from severe straggler issues in airborne SAR networks under dynamic and unstable communication conditions. To address this challenge, we propose AsyncFed-SAR, an efficient asynchronous FL framework tailored for SAR image target recognition. Specifically, AsyncFed-SAR adopts an adaptive online sensor selection (OSS) strategy inspired by reinforcement learning principles. This method designs a novel model quality score as a reward and introduces a virtual queue to ensure fairness, thereby maintaining robust performance even in communication-constrained environments. To further mitigate model staleness introduced by asynchronous updates, we propose a gradient delay compensation (GDC) strategy, which approximates the stale gradient contributions through a lightweight Hessian estimation based on the gradient outer product. Theoretical analysis provides convergence guarantees and demonstrates the framework's capability to reduce total training time. Extensive experiments on the MSTAR and FUSAR-Ship datasets validate that AsyncFed-SAR significantly outperforms baseline methods in generalization and robustness while achieving lower training time and faster convergence.