Cryptographic Fingerprinting for Network Devices Based on Triplet Network and Fuzzy Extractors

Device fingerprinting is a key technology in cybersecurity, which enables organizations to identify potential vulnerabilities, gain valuable insights into their network infrastructure, and enhance overall defense mechanisms. However, the complexity and dynamics of cyberspace make it extremely challenging to generate unique, robust, and tamper-resistant device fingerprints. In this paper, we propose a cryptographic device fingerprinting framework for network devices, which utilizes triplet network to cluster the feature information of data samples into embeddings, and then apply fuzzy extractors to generate a cryptographic fingerprint for each device based on the feature information in each cluster. In order to overcome the discontinuity of embeddings in Hamming space output by triples networks, which degrades the robustness of fingerprints, we use gray code to transform embeddings before applying fuzzy extractors. The experimental results show that the method proposed can obtain unique and robust fingerprint encoding for the same type of device in noisy environments, and supports incremental fingerprint encoding for newly added devices through a small number of sample learning. The experimental results show that the classification accuracy reaches 99.99%, and the histogram of generated fingerprints conform to the Gaussian distribution, which reflects the excellent cryptographic characteristics.