Entropy aided RF-DNA fingerprint learning from Gabor-based images

The number of devices connected to the internet have been increasing and shape Internet of Things (IoT). The security of IoT is an issue due to the use of weak or no encryption. Specific Emitter Identification (SEI) was introduced to overcome this issue by introduce RF-DNA fingerprinting exploring the PHY layer features. Recently, The SEI performance improved by the usage of the signal’s Time Frequency (TF) representation and accelerated using the Deep learning (DL) Convolutional Neural Network (CNN). While the classification accuracy has been improved from using raw signals learning the amount of data generated is large and computationally expensive. This work investigate the usage of statistical thresholds like entropy applied to ”tiles” selected from the signals’ TF representation to reduce the amount of data generated. The results show that the entropy based data reduction decrease the average classification accuracy by 0.86% compared to the usage of the full gray-scale image at 30dB. The usage of enhanced tiles selection algorithms shows an improvement in the average classification accuracy by 25% from the original tile selection procedure at 9dB.

Grantor dc:publisher

University of Tennessee at Chattanooga

Year dc:date.available

2024

Author dc:creator

• Taha, Mohamed Alfatih

Contributors dc:contributor

• Reising, Donald R.
• Loveless, Thomas Daniel; Fadul, Mohamed M. K.
• College of Engineering and Computer Science

• Internet of things
• Deep learning (Machine learning)
• Computer security

Licence

• http://rightsstatements.org/vocab/InC/1.0/

Language dc:language

English, eng

Repository record dc:identifier

https://scholar.utc.edu/theses/795

OAI identifier oai:identifier

oai:scholar.utc.edu:theses-1974