Doubling Down on AI

One of the biggest challenges that face autonomous vehicle manufacturers is the risk of hackers tampering with the artificial intelligence (AI) algorithms designed to recognize road signs and other information that these systems collect from their surroundings. But a different type of threat that alters the data coming into the system can be equally dangerous.

For example, in spite of standardization efforts, it may be possible to alter the content of a traffic sign with a few strategically placed pieces of tape. Even in human-operated vehicles, where onboard systems are relied upon to present information or control certain driving functions, such tampering could present a hazard. To remedy this type of low-tech hacking, a second layer of AI processing has been developed that is based on computer vision. It analyzes objects from a second perspective and compares the two. If a discrepancy exists, it can signal that someone has attempted to hack the system.

This is just one example of how important it is to ensure that, when it comes to machine learning and AI systems, the data as well as the algorithms are protected against manipulation.