Tutorial: Deep Learning for RF Signal Classification and Fingerprinting

Description

Approved for 3 CompTIA CEUs: A+, Network+

Due to the unprecedented scale of connected devices today, designing scalable, accurate, energy-efficient and tamper-proof authentication mechanisms has now become more important than ever. Moreover, with reconfigurable radios, there is a rich diversity of protocols and standards, modulation schemes, and other dynamic transmitter-side settings that cannot be decided ahead of deployment. Drawing from hands-on experiences as a performer in the DARPA Radio Frequency Machine Learning Systems (RFMLS) program, this tutorial presents practical methodologies of extracting "RF fingerprints", i.e., unique signatures embedded in the transmitted signals arising from unchangeable and device-specific hardware characteristics. Using pre-developed code blocks, the tutorial will provide a hands-on experience and thorough coverage of (i) the causes of such fingerprints using experimental results from software defined radios, (ii) deep convolutional neural network (CNN) architectures that are effective at learning such fingerprints, (iii) feature engineering and signal pre-processing steps that increase classification accuracy, (iv) transfer learning techniques where CNNs designed for unrelated domains (such as image classification) are retrained for signal classification. The tutorial will also cover issues related to scalability (50 devices vs. 10K devices) and associated processing overheads, with an analysis of tradeoffs that should be considered when porting the architectures in real-time systems and resource-constrained devices. Finally, the tutorial will provide insights on how these fingerprinting with deep CNNs techniques can be used for other applications such as covert channel design, the relatively simpler problem of modulation classification, address-free IoT, as well as "train once, deploy anywhere" channel-independent device classification.

Tracks:

• 
  Continuing Education
• 
  Technical Tutorials