Flashing Lights and Autonomous Vehicles: Researchers Uncover a Surprising Risk
In a recent episode of the Security Now podcast, hosts Steve Gibson and Leo Laporte discussed a fascinating study that reveals a new safety concern for self-driving cars. The research, conducted by a team from Ben-Gurion University of the Negev and the Japanese technology firm Fujitsu, found that flashing lights from emergency vehicles can cause "digital epileptic seizures" in the image-based autonomous driving systems used by self-driving cars.
The study was inspired by reports of Tesla vehicles using the Autopilot feature crashing into stationary emergency vehicles between 2018 and 2021. The researchers hypothesized that the flashing lights of the emergency vehicles could be the culprit behind these incidents.
To test their theory, the team used five off-the-shelf dashboard cameras as a proxy for the image-based autonomous driving systems. They exposed these cameras to flashing light patterns similar to those used by emergency vehicles and found that the object detection algorithms became less accurate, potentially leading to crashes.
Steve and Leo delved into the technical details of the study, explaining how the researchers used a software fix to mitigate the issue by training the object detection algorithms to specifically recognize vehicles with flashing lights. However, they also discussed the limitations of the study, as the researchers could not test their findings on actual self-driving car systems due to the proprietary nature of the technology.
The conversation also touched on the broader implications of this research for the development and regulation of self-driving cars. Leo raised the point that some automakers might be moving too quickly to deploy these technologies without sufficient testing and validation. Steve emphasized the importance of independent research like this study in uncovering potential safety issues and informing the development of more robust and reliable autonomous driving systems.
Both hosts agreed that the study highlights the need for a more collaborative approach between automakers, researchers, and regulators to ensure the safety of self-driving cars. They also stressed the importance of continued research and testing to identify and address potential risks associated with these rapidly evolving technologies.
The self-driving car segment of the Security Now podcast offers a fascinating look at an unexpected safety concern for autonomous vehicles. It also serves as a reminder of the complex challenges that must be addressed as we move towards a future with more self-driving cars on our roads.
To learn more about this research and its implications, as well as other exciting topics like Microsoft's AI training practices, the Tor network's call for help, and the incredible journey of Voyager 1, be sure to listen to the full episode of Security Now.
