Tutorial: Adversarial Radio Frequency Machine Learning (RFML) Using PyTorch

Description

Approved for 3 CompTIA CEUs: A+, Network+

The concept of deep learning has revitalized machine learning research in recent years. In particular, researchers have demonstrated the use of deep learning for a multitude of tasks in wireless communications, such as signal classification, waveform creation, and cognitive radio, colloquially coined Radio Frequency Machine Learning (RFML) by the Defense Advanced Research Projects Agency (DARPA). Traditionally, these deep learning solutions are developed using static datasets or require the interfacing of a digital signal processing framework, such as GNU Radio or LiquidDSP, with a machine learning framework, such as Keras or PyTorch. Given these facts, this tutorial aims to: introduce the attendee to the state-of-the-art in RFML research for military and commercial applications, describe how to leverage the PyTorch toolkit for creating their own deep learning solutions, show how these solutions can be vulnerable to adversarial machine learning techniques, and, ultimately, how to harden these deep learning solutions against attacks. In addition, advanced PyTorch concepts will be introduced such as: data augmentation in the training loop allowing for faster and more robust training, arbitrary channel modeling allowing to directly optimize waveforms for a given channel model, and optimization of wireless communications specific objective functions.

Tracks:

• 
  Continuing Education
• 
  Technical Tutorials