COMBAT: COntinuous Monitoring of Biosignals for AuThenticating SWaP Defense Equipment

The rapid proliferation of Internet of Things (IoT) devices, including wearable technologies such as smartwatches and portable sensors, has introduced significant challenges in maintaining security, data integrity, and confidentiality. The integration and coexistence of third-party untrusted vendor devices within real-world networks exacerbate these challenges, as a single compromised IoT device can render an entire network vulnerable to cyberattacks. In defense environments, where sensitive data and equipment are at stake, the need for robust, continuous, and scalable authentication mechanisms has become paramount. While effective in some scenarios, traditional Identity Credential and Access Management (ICAM) techniques suffer from challenges such as computational overhead, vulnerability to spoofing, limited scalability, and user inconvenience. This SBIR Phase II project addresses these gaps by developing an innovative continuous Multi-Factor Authentication (MFA) system tailored for defense applications. The proposed solution combines advanced biometric technologies and behavioral analytics to provide a lightweight, user-friendly, and energy-efficient design capable of operating in dynamic and challenging defense environments. The system will non-intrusively collect multiple biosignals, including gait, Electrocardiogram (ECG), Electroencephalogram (EEG), Electromyogram (EMG), and Galvanic Skin Response (GSR), enabling comprehensive and secure identity verification. We will leverage novel pre-processing techniques alongside a dynamic and lightweight transformer-based network to process these biosignals efficiently. Unlike conventional machine learning models such as Convolutional Neural Networks (CNNs) and Deep Neural Networks (DNNs), the proposed approach will employ early-exit-based transformer architectures to optimize performance and reduce computational overhead. A key innovation of this system is its ability to dynamically select relevant biosignals for authentication, thereby preventing redundancy and minimizing energy consumption. The pre-processing pipeline will be designed to adapt to real-world defense operating conditions, ensuring robustness against extreme personnel movements such as walking, running, or jumping, while maintaining low false positive rates. By the conclusion of this project, we aim to deliver a prototype with a higher Technology Readiness Level (TRL), demonstrating real-world applicability in defense settings. Beyond defense applications, the proposed MFA system will be adapted for commercial use, enabling secure and continuous authentication for enterprise and consumer IoT devices. This work represents a significant advancement in the field of IoT security, addressing critical challenges posed by the increasing sophistication of cyber threats while meeting the unique requirements of defense environments.