Cybersecurity in Digital Twins of Electric Vehicle’s LIBs: Unveiling a Robust TTB-GA Attack

Abstract

Virtual replicas of physical systems, known as Digital Twins (DT), can offer innovative solutions for optimizing and forecasting battery management systems (BMS). However, their security remains a major concern. A new type of attack called Time Tampering Black-Box Attack Genetic Algorithm (TTB-GA) is introduced in this paper to study security in DT Intelligent Transportation Systems (DT-ITS). TTB-GA exploits the sensitivity of time series data and effectively deceives prediction models by altering input data’s timing within realistic ranges. To enhance the efficiency of locating and querying, customized operators such as mutation and fitness are designed within the GA-based search framework. Our attack achieves a remarkable success rate of 98% for Long short-term memory (LSTM) and 96% for Gated Recurrent Unit (GRU) models, exposing a critical vulnerability in digital twin security. Furthermore, we demonstrate the limitations of a distributed detection scheme combining an Autoencoder, a Convolutional Neural Network (CNN), and an Extended Kalman Filter (EKF), emphasizing the need for a robust and adaptive defenses. By exposing a novel and highly effective attack method (TTB-GA) targeting temporal vulnerabilities in time series data, and emphasizing the limitations of existing defense mechanisms against such attacks, our research significantly contributes to digital twin security.