原文信息:
False data injection attacks on data markets for electric vehicle charging stations
原文链接:
https://www.sciencedirect.com/science/article/pii/S2666792422000166
Abstract
Modern societies use machine learning techniques to support complex decision-making processes (e.g., renewable energy and power demand forecasting in energy systems). Data fuels these techniques, so the quality of the data fed into them determines the accuracy of the results. While the amount of data is increasing with the adoption of internet-of-things, most of it is still private. Availability of data limits the application of machine learning. Scientists and industry pioneers are proposing a model that relies on the economics of data markets, where private data can be traded for a price. Cybersecurity analyses of such markets are lacking. In this context, our study makes two contributions. First, it designs a data market for electric vehicle charging stations, which aims to improve the accuracy of electric vehicle charging demand forecasts. Accurate demand forecasts are essential for sustainable operations of the electric vehicle - charging station - power grid ecosystem, which, in turn, facilitates the electrification and decarbonization of the transportation sector. On the other hand, erroneous demand forecasts caused by malicious cyberattacks impose operational challenges to the ecosystem. Thus, the second contribution of our study is to examine the feasibility of false data injection attacks on the data market for electric vehicle charging stations and to propose a defense mechanism against such attacks. We illustrate our results using data from electric vehicle charging stations in Manhattan, New York. We demonstrate that the data market improves forecasting accuracy of charging stations and reduces the effectiveness of false data injection attacks. The purpose of this work is not only to inform electric vehicle charging stations about the economic benefits of data markets, but to promote cyber awareness among data market pioneers and stakeholders.
Keywords
Data markets
Demand forecasts
Electric vehicle charging stations
Kullback-Leibler divergence
Machine learning
Quantile linear regression
Graphics
Fig. 1. Operation modes (1–6) of EVCS data market and compromised data owners under a false data injection cyberattack on the data market.
Fig. 2. Computations of the data market operator.
Fig. 3. (a) Training and (b) Test errors, based on the Pinball loss in Eq. (4b), of the EVCS demand forecast models. The models have 𝛼 = 0.5.
Fig. 4. (a) Difffference in EVCS demand forecasts using public+private data 𝛼 =0.5, (b) EVCS demand forecasts using public+private data 𝛼 = [0.1, 0.5, 0.9].
Fig. 5. Shapley values and the data payments (in cents) of the data features used for forecasting the EVCS demand in Fig 4(a). The data payments are scaled down by 1000 to visualize them and Shapley values together.
Fig. 6. False data injection attack on the data market to reduce the EVCS demand forecast by 10% at 𝑡 = 13, launched by compromising EV 𝑆𝑜𝐶 data.
Fig. 7. False data injection attack on the data market to reduce the EVCS demand forecast by 10% at 𝑡 =13, by compromising data of neighborhood EVCSs.
关于Applied Energy
本期小编:杜莹; 审核人:郭加澄
《Applied Energy》是世界能源领域著名学术期刊,在全球出版巨头爱思唯尔 (Elsevier) 旗下,1975年创刊,影响因子11.446,CiteScore 20.4,高被引论文ESI全球工程期刊排名第4,谷歌学术全球学术期刊第50,本刊旨在为清洁能源转换技术、能源过程和系统优化、能源效率、智慧能源、环境污染物及温室气体减排、能源与其他学科交叉融合、以及能源可持续发展等领域提供交流分享和合作的平台。开源(Open Access)姊妹新刊《Advances in Applied Energy》现已正式上线。在《Applied Energy》的成功经验基础上,致力于发表应用能源领域顶尖科研成果,并为广大科研人员提供一个快速权威的学术交流和发表平台,欢迎关注!
公众号团队小编招募长期开放,欢迎发送自我简介(含教育背景、研究方向等内容)至wechat@applied-energy.org
点击“阅读原文”
提交文章吧
喜欢我们的内容?
点个“赞”或者“再看”支持下吧!