BATON ROUGE, La., April 25, 2024 (GLOBE NEWSWIRE) -- Since the pandemic, our primary means of interacting have evolved to include video conferencing applications, which are broadly used to connect geographically distant people for work, school, and even socially. During these interactions, it’s not uncommon for a user to turn off his or her microphone and/or camera out of concern for privacy. However, microphones and cameras can still leak other kinds of information beyond what is seen and heard through “micro signals,” which are too tiny for humans to recognize but detectable by machines.
Understanding and mitigating the risk these signals present is at the heart of LSU Computer Science Assistant Professor Chen Wang’s research, which was just funded by a National Science Foundation (NSF) CAREER Award.
“Any signals that are human-unobservable but machine-distinguishable are micro signals,” Wang said. “In online meetings, micro signals are the tiny audio and video signals, which are dominated by human speech and camera scenes and thus, usually unnoticed. We find that the audio feedback (i.e., echo sounds) in online meetings can bring location-dependent information back, which can be exploited to locate the user.
“So, when you turn off your camera or use a virtual background to hide your location, if your mic is on, your location may still be leaked. Moreover, we find that a video camera may capture your activities outside of its view angles. This is because the monitor’s screen lights and their reflections can be used for sensing the surroundings. Then, your secret actions in the camera’s blind spots, such as typing a password, may still be ‘seen’ by other meeting participants.”
An attacker can send malicious acoustic signals remotely, which sense the user’s physical surroundings and return to the attacker with location-specific echo signals. Deep-learning algorithms developed by the attacker are then used to circumvent the echo cancellation mechanisms enforced by audio streaming systems, maximizing the retrieval of sensitive echoes.
Wang’s research group will combat these attacks with targeted defenses that address their weaknesses separately, as well as broader techniques that cast a wider net.
“For example, we aim to detect suspicious echo sounds in online meetings and filter out these unwanted signals,” he said. “We can also shuffle the key positions to prevent an adversary from exploiting screen light signals to sense the user’s keystrokes off the camera.
“The general micro-signal removal aims to eliminate any micro signals and their variants, even if they are not identified by our study. We propose to develop AI algorithms to remove all micro signals from the audio and video data of online meetings without degrading the user experience.”
One added component of the project is that it will contribute to cybersecurity education through curriculum development, demo platform implementation, graduate/undergraduate student training, K-12 involvement, public outreach, and underrepresented student engagement in research.
On a more personal note, Wang said that the NSF CAREER Award marked a milestone in his career and served as recognition of his potential to serve as an academic role model in cybersecurity research and education.
“This five-year funding support enables me to continue working towards my career goal, ‘exploring unobtrusive sensing for enhanced security and better living and fostering cybersecurity education with hands-on learning,’” Wang said. “I would like to thank the LSU College of Engineering Research Facilitation Office, Provost’s Fund for Innovation Research-Summer Stipend, Louisiana Board of Regents Research Competitiveness Subprogram fund, and all of my lab students. They’ve helped me to achieve my academic success.”
Like us on Facebook (@lsuengineering) or follow us on X (formerly Twitter) and Instagram (@lsuengineering).
###
