The European Union Agency for Cybersecurity (ENISA) and the Joint Research Centre of the European Commission (JRC) have issued a joint guidance on “Cybersecurity Challenges in the Update of Artificial Intelligence in Autonomous Driving.”
- Because autonomous vehicles use AI for many aspects traditionally handled by human drivers, the classic approach of “securing digital systems” is not sufficient. You must consider AI-specific issues in the context of the full supply chain involved in their development and integration with other automotive systems.
- AI and Machine Learning are the keystone of highly-accessorized smart cars and will be the key for the next generation cars.
- The most significant and well-known vulnerabilities and possible attack scenarios on AI models employed in AV’s involve computer vision techniques.
- A crucial challenge is the right handling of edge cases, where an unknown situation outside of the training data distribution is encountered: e.g. ignoring a stop sign partially covered by snow.
The report analyzes common AI vulnerabilities in autonomous vehicles and recommendations to address them.