In September 2024, a significant vulnerability in Apple’s Vision Pro mixed reality headset was revealed, exposing users to potential privacy breaches through an attack known as GAZEploit (CVE-2024-40865). This flaw allowed malicious actors to infer text entered on the device’s virtual keyboard by analyzing eye movements of the virtual avatar used during gaze-controlled typing. Researchers from the University of Florida, CertiK Skyfall Team, and Texas Tech University discovered that the flaw affected the Presence component, enabling unauthorized reconstruction of typed text based on avatar gaze patterns.
The GAZEploit attack leverages the vulnerability inherent in the gaze-controlled text entry system when users share virtual avatars. By capturing and analyzing video of the avatar, attackers could discern what the user was typing on the virtual keyboard. This sophisticated attack utilized a supervised learning model trained on Persona recordings, eye aspect ratio (EAR), and gaze estimation to map gaze directions to specific keys on the virtual keyboard, allowing for precise keystroke inference.
Apple responded swiftly to the discovery, addressing the vulnerability with a patch included in visionOS 1.3, released on July 29, 2024. The company resolved the issue by suspending the Persona component when the virtual keyboard is active, effectively mitigating the risk of keystroke inference. This fix was a crucial step in enhancing user privacy and securing the mixed reality environment against such threats.
The GAZEploit attack highlights the importance of robust privacy protections in emerging technologies, particularly those involving gaze-controlled inputs. As mixed reality systems become more prevalent, it is essential for developers and manufacturers to implement comprehensive security measures to safeguard user data and prevent potential exploitation. This incident serves as a reminder of the evolving nature of cyber threats and the need for continuous vigilance and innovation in cybersecurity.
