Using Biometric Data to Authenticate Tactical Edge Network Users

The Internet of Things (IoT) is impacting several areas. Using sensors and actuators in different contexts, such as smart cities, industry 4.0, healthcare, and agriculture, creates new interactions that can only occur due to connecting devices. As IoT devices join the military domain, new problems arise, like security and authentication in unreliable networks. This article proposes a novel approach for continuous user authentication to improve wearable device security in the Internet of Battle Things (IoBT) context. The proposed approach uses a Recurrent Neural Network (RNN) to directly optimize the embedding using the triplet loss. The created embedding represents the authorized user gait as a n-dimensional vector and enables comparison with other users’ gait using a \(L_2\) distance corresponding to a measure of gait similarity. The proposed system achieves a 92.12% rank-1 accuracy on user identification and a 12.33% equal error rate for the user validation task.