DeepFake: Deep Dueling-based Deception Strategy to Defeat Reactive Jammers

Abstract

In this paper, we introduce DeepFake, a novel deep reinforcement
\nlearning-based deception strategy to deal with reactive jamming attacks. In
\nparticular, for a smart and reactive jamming attack, the jammer is able to
\nsense the channel and attack the channel if it detects communications from the
\nlegitimate transmitter. To deal with such attacks, we propose an intelligent
\ndeception strategy which allows the legitimate transmitter to transmit "fake"
\nsignals to attract the jammer. Then, if the jammer attacks the channel, the
\ntransmitter can leverage the strong jamming signals to transmit data by using
\nambient backscatter communication technology or harvest energy from the strong
\njamming signals for future use. By doing so, we can not only undermine the
\nattack ability of the jammer, but also utilize jamming signals to improve the
\nsystem performance. To effectively learn from and adapt to the dynamic and
\nuncertainty of jamming attacks, we develop a novel deep reinforcement learning
\nalgorithm using the deep dueling neural network architecture to obtain the
\noptimal policy with thousand times faster than those of the conventional
\nreinforcement algorithms. Extensive simulation results reveal that our proposed
\nDeepFake framework is superior to other anti-jamming strategies in terms of
\nthroughput, packet loss, and learning rate.

Topics & Concepts