Towards Adversarial and Unintentional Collisions Detection Using Deep Learning
Hai Ngoc Nguyen, Tien Dang Vo-Huu, Triet Dang Vo-Huu, Guevara Noubir · 2019
We introduce a set of techniques to achieve transfer learning from computer vision to RF spectrum analysis. In this paper, we demonstrate the usefulness of this approach to scale the learning, accuracy, and efficiency of detection of adversarial and unintentional communications collisions using VGG-16. We achieve high accuracy (94% collisions detected) on a DARPA Spectrum Collaboration Challenge (SC2) dataset.