arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2303.03535eess.SYcs.SY
收藏

电动汽车充电控制中面向特定目的的去中心化算法网络攻击探索

Exploration of For-Purpose Decentralized Algorithmic Cyber Attacks in EV Charging Control

Mahan Fakouri Fard, Xiang Huo, Mingxi Liu

AI总结:

本文首次研究电动汽车充电控制中由参与智能体发起的面向特定目的的算法攻击,设计隐蔽与非隐蔽攻击向量,并通过仿真验证其可行性与对配电网的影响。

摘要:

分布式与去中心化多智能体优化(DMAO)算法能够控制大规模电网边缘资源(如电动汽车 EV),以提供电网服务。尽管 DMAO 具有出色的可扩展性,但由于其高度依赖频繁的点对点通信,本质上容易遭受网络攻击。现有相关网络安全研究主要关注旨在破坏整个控制系统的广谱攻击,从而丧失了实现特定攻击目的的可能性。本文首次探索了由参与智能体发起并与 DMAO 交互以实现自利攻击目的的新型面向特定目的的算法攻击。以去中心化电动汽车充电控制问题作为说明性用例,设计了具有和不具有隐蔽特性的理论面向特定目的的攻击向量。电动汽车充电控制仿真结果表明,所提出的算法性面向特定目的攻击具有可行性,并展示了此类攻击对配电网的影响。

英文摘要:

Distributed and decentralized multi-agent optimization (DMAO) algorithms enable the control of large-scale grid-edge resources, such as electric vehicles (EV), to provide power grid services. Despite its great scalability, DMAO is fundamentally prone to cyber attacks as it is highly dependent on frequent peer-to-peer communications. Existing cyber-security research in this regard mainly focuses on \emph{broad-spectrum} attacks aiming at jeopardizing the entire control system while losing the possibility of achieving specific attacking purposes. This paper, for the first time, explores novel \emph{for-purpose} algorithmic attacks that are launched by participating agents and interface with DMAO to achieve self-interest attack purposes. A decentralized EV charging control problem is formulated as an illustrative use case. Theoretical \emph{for-purpose} attack vectors with and without the stealthy feature are devised. Simulations on EV charging control show the practicability of the proposed algorithmic \emph{for-purpose} attacks and the impacts of such attacks on distribution networks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑