With the integration of artificial intelligence (AI) in the auto industry, self-driving cars offer exciting possibilities for semi-autonomous driving. However, this advancement raises concerns about cybersecurity and protecting drivers’ personal information. As automakers explore the potential of AI in vehicles, they must address the challenge of safeguarding data and ensuring privacy.
Dr. M. Hadi Amini, an assistant professor at FIU’s College of Engineering and Computing, is leading research on AI integration in transportation cybersecurity and resiliency, funded by the U.S. Department of Transportation. His work focuses on a decentralized form of AI, federated learning, to protect drivers’ data and enhance overall computing efficiency.
AI and vulnerabilities in modern cars
As vehicles become more computerized, they also become susceptible to cyberattacks and potential privacy leaks. Demonstrations by ethical hackers have shown that modern car technology, such as infotainment systems, can be compromised. This cybersecurity concern has prompted researchers to explore AI solutions that can protect data and ensure the secure operation of transportation systems.
Addressing privacy concerns
One of the primary concerns for automakers is the storage of drivers’ personal information. AI algorithms require extensive data for learning and decision-making, including sensitive details like phone contacts, location data, and garage door codes. If a central server in a network of cars is hacked, it could compromise the personal information of all drivers in that network. Protecting privacy becomes a critical challenge for AI integration in autonomous vehicles.
Dr. Amini’s research focuses on federated learning, a decentralized form of AI that lessens the reliance on a central server. Instead of aggregating all data at a central location, federated learning allows individual cars to process and learn from their data. Cars then transmit algorithm suggestions, devoid of raw data, to servers that enhance the overall algorithm for the entire network. This approach safeguards drivers’ privacy and offers efficient and scalable computing for an increasing number of cars.
The potential of federated learning
Federated learning addresses the vulnerability of centralized machine learning, where a failure in the central server can bring down the entire system. In contrast, a distributed machine learning approach enables the rest of the system to function for some time, relying on local data, even during an attack or disaster. By adopting federated learning, automakers can capitalize on AI advancements while minimizing the risk of data breaches and ensuring secure transportation systems.
While no system can ever be entirely secure, federated learning provides a promising pathway for the auto industry. By preserving drivers’ privacy and decentralizing AI computation, automakers can harness AI’s potential without compromising their customers’ safety and data privacy.
Integrating AI presents exciting opportunities and critical challenges as self-driving cars become a reality. Ensuring data privacy and protecting against cyberattacks are top priorities for automakers. Dr. M. Hadi Amini’s research on federated learning offers a potential solution to these challenges, allowing for decentralized AI computation and safeguarding drivers’ personal information. With a focus on responsible and secure AI integration, automakers can unlock AI’s full potential while ensuring drivers’ safety and privacy in the age of autonomous vehicles.
