Shaked Adi, Dvir Elshich, and Adar Peleg, students at the faculty, will present a vulnerability they discovered in Google's protocol at the SecTor 2025 conference in Canada, an important international conference held as part of Black Hat.
The vulnerability was discovered as part of a project the three conducted under the direction of Prof. Avi Mendelsohn, Rom Himmelstein, Amit Levy and Stav Cohen as part of the ATLAS AI research lab at the Taub Faculty of Computer Science.
The three undergraduate students joined the lab only three months ago, and since then they have not only studied the subject of deep learning models but have also managed to discover the same critical vulnerability. This is a vulnerability in A2A, Google's new communication protocol for artificial intelligence agents, which was launched only a month ago. This vulnerability allows an external user to steal information, implant malicious code and control artificial intelligence agents. According to the students, this is a living example of the risks that exist in artificial intelligence systems and the importance of academia in building Israel as a power in the field.
The students informed Google of the vulnerability, which they will present at the SecTor conference. They note that this is not the only weakness in the system, and therefore they will continue this research.
[Back to the news index]