This report evaluates the data quality of two standard publicly available automotive intrusion detection system datasets: the SOME/IP Attack Dataset and the Survival Analysis Dataset (SAD). The data quality evaluation is guided by the data quality model of the comprehensive ISO/IEC 5259 data quality standard series and incorporates domain specific requirements and data usage context relevant to automotive networks. The automotive network data quality requirements are aligned with the AI Act Article 10 provisions on data and data governance requirements, which emphasizes the importance of trustworthy data subject for AI model training. The report presents key findings and reflections on both datasets to enhance understanding, ensure compliance, and support their adoption in the development and validation of AI/ML-based automotive intrusion detection systems.
This work is supported by the Vinnova INTelligent sEcuRity SoluTIons for Connected vEhicles (INTERSTICE) project (reference number: 2024-00661). This work is also partially supported by the EU project Citcom.AI, one of the EU's four AI TEFs (Testing and Experimental Facilities for Smart and Sustainable Cities and Societies).