Zero-Bias Deep Learning for Accurate Identification of Internet-of-Things (IoT) Devices

The Internet of Things (IoT) provides applications and services that would\notherwise not be possible. However, the open nature of IoT make it vulnerable\nto cybersecurity threats. Especially, identity spoofing attacks, where an\nadversary passively listens to existing radio communications and then mimic the\nidentity of legitimate devices to conduct malicious activities. Existing\nsolutions employ cryptographic signatures to verify the trustworthiness of\nreceived information. In prevalent IoT, secret keys for cryptography can\npotentially be disclosed and disable the verification mechanism.\nNon-cryptographic device verification is needed to ensure trustworthy IoT. In\nthis paper, we propose an enhanced deep learning framework for IoT device\nidentification using physical layer signals. Specifically, we enable our\nframework to report unseen IoT devices and introduce the zero-bias layer to\ndeep neural networks to increase robustness and interpretability. We have\nevaluated the effectiveness of the proposed framework using real data from\nADS-B (Automatic Dependent Surveillance-Broadcast), an application of IoT in\naviation. The proposed framework has the potential to be applied to accurate\nidentification of IoT devices in a variety of IoT applications and services.\nCodes and data are available in IEEE Dataport.\n