A Comprehensive Analysis of Datasets for Automotive Intrusion Detection Systems

• DOI: 10.32604/cmc.2023.039583
• Publication Date: 2023-10-09T09:29:23Z
• Authors: Seyoung Lee; Wonsuk Choi; Insup Kim; Ganggyu Lee; Dong Hoon Lee
• Keywords: Adaptability

Recently, automotive intrusion detection systems (IDSs) have emerged as promising defense approaches to counter attacks on in-vehicle networks (IVNs). However, the effectiveness of IDSs relies heavily quality datasets used for training and evaluation. Despite availability several IDSs, there has been a lack comprehensive analysis focusing assessing these datasets. This paper aims address need dataset assessment in context IDSs. It proposes qualitative quantitative metrics that are independent specific evaluate These take into consideration various aspects such description, collection environment, attack complexity. evaluates eight commonly using proposed metrics. The evaluation reveals biases datasets, particularly terms limited contexts diversity. Additionally, it highlights were mostly injected without considering normal behaviors, which poses challenges evaluating machine learning-based emphasizes importance addressing identified limitations existing improve performance adaptability can serve valuable guidelines researchers practitioners selecting constructing high-quality security applications. Finally, this presents requirements including representativeness, diversity, balance.