Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Behavior-Aware Efficient Detection of Malicious EVs in V2G Systems

Ruixiang Wu, Xudong Wang, Tongxin Li

TL;DR

A safety-enabled group testing scheme, \ouralg, is proposed, which combines the efficiency of probabilistic group testing with ML predictions and the robustness of combinatorial group testing and is proved that \ouralg is O(d)-consistent and O(d\log n)-robust, striking a near-optimal trade-off.

Abstract

With the rapid development of electric vehicles (EVs) and vehicle-to-grid (V2G) technology, detecting malicious EV drivers is becoming increasingly important for the reliability and efficiency of smart grids. To address this challenge, machine learning (ML) algorithms are employed to predict user behavior and identify patterns of non-cooperation. However, the ML predictions are often untrusted, which can significantly degrade the performance of existing algorithms. In this paper, we propose a safety-enabled group testing scheme, \ouralg, which combines the efficiency of probabilistic group testing with ML predictions and the robustness of combinatorial group testing. We prove that \ouralg is $O(d)$-consistent and $O(d\log n)$-robust, striking a near-optimal trade-off. Experiments on synthetic data and case studies based on \textsc{ACN-Data}, a real-world EV charging dataset validate the efficacy of \ouralg for efficiently detecting malicious users in V2G systems. Our findings contribute to the growing field of algorithms with predictions and provide insights for incorporating distributional ML advice into algorithmic decision-making in energy and transportation-related systems.

Behavior-Aware Efficient Detection of Malicious EVs in V2G Systems

TL;DR

A safety-enabled group testing scheme, \ouralg, is proposed, which combines the efficiency of probabilistic group testing with ML predictions and the robustness of combinatorial group testing and is proved that \ouralg is O(d)-consistent and O(d\log n)-robust, striking a near-optimal trade-off.

Abstract

With the rapid development of electric vehicles (EVs) and vehicle-to-grid (V2G) technology, detecting malicious EV drivers is becoming increasingly important for the reliability and efficiency of smart grids. To address this challenge, machine learning (ML) algorithms are employed to predict user behavior and identify patterns of non-cooperation. However, the ML predictions are often untrusted, which can significantly degrade the performance of existing algorithms. In this paper, we propose a safety-enabled group testing scheme, \ouralg, which combines the efficiency of probabilistic group testing with ML predictions and the robustness of combinatorial group testing. We prove that \ouralg is -consistent and -robust, striking a near-optimal trade-off. Experiments on synthetic data and case studies based on \textsc{ACN-Data}, a real-world EV charging dataset validate the efficacy of \ouralg for efficiently detecting malicious users in V2G systems. Our findings contribute to the growing field of algorithms with predictions and provide insights for incorporating distributional ML advice into algorithmic decision-making in energy and transportation-related systems.

Paper Structure

This paper contains 22 sections, 8 theorems, 17 equations, 7 figures, 1 table, 3 algorithms.

Table of Contents

  1. Introduction
  2. Previous Work
  3. Contributions
  4. Problem Formulation
  5. Untrusted ML Predictions
  6. Performance Benchmarks
  7. Detecting Malicious Users Efficiently from Behavioral Data
  8. Warm-Up: Why Don't Trust All Predictions?
  9. Our Solution: Incorporating a Safety Threshold
  10. Theoretical Results
  11. Expected Regret Analysis
  12. Best-of-Both-Worlds Performance of GTUA
  13. Conclusion
  14. Probabilistic and Combinatorial Group Testing Procedures
  15. Useful Lemmas
  16. ...and 7 more sections

Key Result

Lemma 1

When trusting the ML predictions $\widetilde{\mathbf{p}}$, the expected number of tests used by the LA satisfies

Figures (7)

  • Figure 1: Illustration of Malicious User Detection in a V2G System (see the model defined in Section \ref{['sec:model']}). A V2G manager is equipped with a sensor to detect malicious EVs from a subset of EVs specified by an algorithm. Machine learning models learned from grid data generate distributional advice $\widetilde{p}$ (denoting the probabilities of behaving maliciously). An algorithm, denoted by GTUA is presented in Section \ref{['sec:GT+MLPred']} for efficient malicious EV detection.
  • Figure 2: Comparison of the average number of tests for LApriorstats, GBSgbsa, and the proposed GTUA (Algorithm \ref{['alg:pseudocode']}) in this work. Shadow areas depict the magnitudes of standard deviations.
  • Figure 3: Comparison of model distributions with actual collected data.
  • Figure 4: Comparison of number of tests each hour using GTUA and the number of EVs to be tested each hour.
  • Figure : Group Testing with Untrusted ML Advice (GTUA)
  • ...and 2 more figures

Theorems & Definitions (14)

  • Definition 1: Tightness Gap
  • Definition 2: Expected $\kappa$-Regret
  • Lemma 1: Generalization of Theorem 1 in priorstats
  • Theorem 1: Expected $\kappa({\textsf{LA}}\xspace)$-Regret
  • Definition 3: Consistency and robustness
  • Corollary 1
  • Theorem 2
  • Lemma 2: Theorem 1 in priorstats
  • Lemma 3: Theorem 2 in priorstats
  • Lemma 4
  • ...and 4 more