Conducted Emission signal-based identification and real-time hardware security with deep learning


Sakacı F. H., YILDIRIM T.

Engineering Applications of Artificial Intelligence, cilt.136, 2024 (SCI-Expanded) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 136
  • Basım Tarihi: 2024
  • Doi Numarası: 10.1016/j.engappai.2024.109025
  • Dergi Adı: Engineering Applications of Artificial Intelligence
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aerospace Database, Applied Science & Technology Source, Communication Abstracts, Compendex, Computer & Applied Sciences, INSPEC, Metadex, Civil Engineering Abstracts
  • Anahtar Kelimeler: Deep learning, Device identification, Electromagnetic emission, Hardware security, Imitation attack protection
  • Yıldız Teknik Üniversitesi Adresli: Evet

Özet

In advancing electronic systems, the replicable nature of devices used in critical environments raises concerns regarding system security. It is of great importance to establish security by identifying electronic devices and computers based on their unique electrical signals, rather than assigning an artificial identity number for security purposes. This paper presents a methodology for identifying electronic devices by capturing their Conducted Emission signal over cables using a Line Impedance Stabilization Network (LISN), followed by applying deep learning techniques. After the extraction of the identities specific to each device based on their electronic characteristics, the establishment of security has been accomplished through these characteristics. The identification process was analyzed through theoretical approaches, and signals were examined by acquiring simulation-based data. After performing the analyses, the data was converted into spectrograms for training the deep learning model using the Conducted Emission signals, and a Convolutional Neural Network (CNN) was used due to its recognized efficacy in handling two-dimensional data formats like spectrogram for the deep learning algorithm. It was observed that with an accuracy of 99.9%, identification was achieved in deep learning results utilizing the data obtained through simulations. Real-time data from 15 devices was collected after training the system with simulation-based data, demonstrating a 99.8% accuracy in accurately identifying the devices. Furthermore, the developed deep learning model was deployed on an Embedded Linux edge device, tested within the system for one day, and achieved an average accuracy of 99.54% in accurately identifying registered devices.