Detection of Distributed Denial of Charge (DDoC) Attacks Using Deep Neural Networks With Vector Embedding

To prevent excessive strain on the electrical grid and avoid long waiting times of the electric vehicle (EV) at charging stations, charging coordination mechanisms have been implemented. However, there is a potential vulnerability that enable adversaries to launch distributed denial of charge (DDoC) attacks. In these attacks, fake charging requests are sent to book charging time slots without showing up for charging. Existing mechanisms assume the requests from EVs are valid and do not address the detection of DDoC attacks. This research paper aims to assess the disruptive capabilities of DDoC attacks on charging coordination mechanisms and utilize deep neural networks incorporated with vector embedding to develop detectors that can protect against these attacks. The detection approach relies on identifying abnormal behavior that deviates from the typical patterns of charging demand at the charging station. To train and evaluate the detectors, we utilize real routes of vehicles and technical parameters of EVs released by their manufacturers to create a benign dataset. Subsequently, various attacks are introduced to generate a malicious dataset. By analyzing this dataset, temporal and spatial correlations are identified, which can be learned by our detectors to detect the attacks accurately. The reason for the design of our detector is based on utilizing the embedding layer to identify concealed patterns in regular charging demand information. Additionally, the deep learning network is employed to comprehend and learn the connections over time in the sequential data, as well as the relationships between the data of adjacent charging stations. We conducted thorough experiments to assess our detectors, and the outcomes demonstrate that the suggested detectors are highly effective in terms of accurately detecting of DDoC attacks while keeping false alarms to a minimum.