Virginia Tech® home


Hack Radio Frequencly Machine Learning Goals

The purpose of the HACK RFML Adversarial Team is to investigate the application of state-of-the-art adversarial machine learning techniques to radio frequency machine learning (RFML) technologies. Adversarial techniques such as evasion attacks, poisoning attacks, and hardware attacks will be investigated. Through investigation of these attack vectors on RFML technologies, the team will also investigate methods by which to harden these technologies against these attacks.

Issues Involved or Addressed

  • Unique challenges of applying state-of-the-art adversarial machine learning techniques from other modalities (images, text, etc.) to the radio frequency machine learning domain
  • Tradeoffs between successful attacks against eavesdroppers vs. successful intended communications
  • Determination of how much knowledge is required about the adversary system for successful attack
  • How to harden our eavesdropper systems against investigated attacks

Methods and Technologies

  • Python and C++ (code development)
  • LiquidDSP (dataset generation)
  • GNU Radio (over-the-air testing)
  • PyTorch (deep learning training, validation, and testing)

Academic Majors of Interest

  • Electrical Engineering
  • Computer Engineering
  • Computer Science
  • Computational Modeling and Data Analytics

Preferred Interests and Preparations

  • Wireless Communications
  • Dataset Creation
  • Machine Learning
  • Adversarial Machine Learning

Team Advisors