Cybersecurity research at the intersection of wearable computing, neural interfaces, and cyber-physical systems security.
Body Machine Interfaces (BMIs) are emerging human-computer interaction systems that translate physiological signals — such as surface electromyography (sEMG), electroencephalography (EEG), and motion data — into control commands for machines, prosthetics, robotic exoskeletons, and other cyber-physical devices. As BMIs become increasingly integrated into medical, rehabilitation, and industrial settings, the security of the monitoring-control loop connecting the human body to downstream actuators becomes a critical concern.
This project investigates the attack surface of BMI systems, with a particular focus on adversarial manipulation of biosignal acquisition, transmission, and processing pipelines. We explore how physical-layer attacks, signal spoofing, and adversarial machine learning techniques can compromise BMI integrity, and develop principled defenses to safeguard patients and users from harm.
Systematically characterize the attack surface of the monitoring-control loop, from biosignal capture through wireless transmission to actuator command execution, using established frameworks such as MITRE ATT&CK.
Investigate whether adversaries can exploit physical proximity or environmental interference to inject malicious signals into sEMG or EEG acquisition pipelines, causing unintended or harmful machine actions.
Evaluate the resilience of deep learning-based gesture and intent recognition models against adversarial perturbations, and develop robust training strategies and anomaly detection mechanisms.
Propose architectural countermeasures — including signal authentication, integrity verification, and safe-mode fallback mechanisms — to harden the end-to-end BMI pipeline against both cyber and physical threats.
The project takes a co-design approach, combining hardware-in-the-loop experimentation with a custom BMI testbed, adversarial machine learning research, and formal security analysis. We use commercial off-the-shelf sEMG sensors and microcontrollers to prototype realistic attack and defense scenarios. Findings are validated through controlled experiments and evaluated against real-world usability constraints.
The project aims to deliver a comprehensive threat taxonomy for BMI systems, a suite of open-source attack simulation tools, and a set of security guidelines and defensive mechanisms suitable for adoption by BMI device manufacturers and healthcare regulators. Results will be disseminated through peer-reviewed publications at top-tier security and human-computer interaction venues.