报告人简介：
叶枫于2019年获得东北大学电气工程及其自动化专业学士学位，2024年获得东南大学控制科学与工程专业博士学位。现为加拿大维多利亚大学电气与计算机工程系博士后研究员。研究方向主要包括分布式协同与优化、智能电网、隐私保护与网络安全。现担任IEEE维多利亚分会执委、IEEE青年工作委员会维多利亚分会主席，以及IEEE Transactions on Vehicular Technology期刊行政编辑（Admin）。
报告摘要：
Differential privacy (DP) is widely used in distributed control to protect sensitive data through noise injection. However, this mechanism also introduces a critical vulnerability: adversaries can launch false noise (FN) attacks by disguising malicious signals as legitimate DP noise, making them highly stealthy and difficult to detect. The problem is further exacerbated in distributed systems with limited observability. In this talk, we address FN attack detection in privacy-preserving distributed control, using distributed energy management systems as a motivating example. We propose a fully distributed, peer-to-peer framework, termed False Noise Attack Detection (FNAD). Each device leverages two-hop neighborhood information and Kalman filtering to estimate neighbor behaviors, and constructs detection indices based on information-theoretic entropy without prior knowledge of attack models. A majority voting mechanism is then employed for robust decision-making. We establish theoretical guarantees on detection performance against representative attacks and validate the effectiveness of FNAD through extensive simulations. This work highlights the fundamental interplay between privacy and security, and provides new insights into designing resilient distributed control systems under privacy constraints.
欢迎广大师生参加！
信息科学与工程学院
人工智能学院（软件学院）
2026年5月11日
