Description

A novel cybersecurity threat named 'WiKI-Eve' has emerged, targeting the security of smartphones connected to modern WiFi routers. This attack, discovered by a team of researchers from universities in China and Singapore, exploits a vulnerability related to Beamforming Feedback Information (BFI), a feature introduced in WiFi 5 (802.11ac) in 2013. BFI allows devices to send their position feedback to routers for more precise signal direction. However, the problem lies in the fact that this data is exchanged in cleartext, making it vulnerable to interception. WiKI-Eve is designed to intercept WiFi signals during password entry, posing a real-time threat that must be executed while the target actively uses their smartphone. Attackers identify the target using an identity indicator like a MAC address, which may require preparatory work such as visual and traffic monitoring. During the attack, the attacker captures the victim's BFI time series using traffic monitoring tools like Wireshark. When the user presses a key, it influences the WiFi antennas behind the screen, generating a distinct WiFi signal. Despite some blurring of keystrokes in the captured data, an algorithm is employed to parse and restore usable information. Moreover, to overcome challenges like typing style and adjacent keystrokes, the researchers employ machine learning, specifically a "1-D Convolutional Neural Network," trained for domain adaptation. A "Gradient Reversal Layer" suppresses domain-specific features, ensuring consistent keystroke recognition. WiKI-Eve's experiments involved participants using different phone models, typing various passwords at different speeds in diverse environments. Results revealed a stable keystroke classification accuracy of 88.9% when employing sparse recovery algorithm and domain adaptation. It achieved an 85% success rate for six-digit numerical passwords in under a hundred attempts, dropping to 23% when the attacker was 10 meters away from the access point. Additionally, the attack successfully deduced passwords for services like WeChat Pay at a rate of 65.8%. In summary, WiKI-Eve showcases how adversaries can extract sensitive data without hacking access points, relying solely on network traffic monitoring and machine learning. To counter this threat, WiFi access points and smartphone apps may need to implement security measures such as keyboard randomization, data traffic encryption, signal obfuscation, CSI scrambling, and WiFi channel scrambling. Protecting against such attacks is crucial to safeguard smartphone users' privacy and security in today's connected world.