A novel jamming strategy-greedy bandit

How Machine Learning Changes the Nature of Cyberattacks on IoT Networks: A Survey

“…Researches relied on that of previous article proves that it is possible to base only on standard rewards for all types of environment (e.g. civilian, military) [151]. They find two new kinds of reward, which are the change of power and the enduring time.…”

“…Also, experiments have demonstrated that the proposed algorithm converges faster to the optimal solution than that used with e-greedy, except when the discriminating parameter of the e-greedy algorithm is optimal (M) [150]. The objective thus becomes finding the optimal discriminant parameter (M) in the first time, in order to apply the e-greedy resolution algorithm in a second step [151]. Thanks to simulation, they compare their results with the 'Jamming bandit'algorithm and show that their method converges faster towards an optimal jamming solution.…”

Apprenticeship learning in cognitive jamming

“…The signaling scheme they use might be binary phase‐shift keying, quadrature phase‐shift keying, multilevel quadrature amplitude modulation, or others, which could be recognized by the jammer that adopts the theory of modulation recognition. For the purpose of denying communications, the jammer aims to learn a jamming strategy using the RL algorithms proposed in other works, and the jamming actions are selected and executed according to the learned policy. When adversarial communicators are experiencing persistent jamming, they must employ some strategies to resist the effects of jamming, such as increasing the power of the transmitted signal, changing signal modulations, switching channels, or adopting antijamming strategies.…”

“…Fortunately, many approaches have also been proposed to learn how to act in an unknown communication environment. The classical theory of reinforcement learning (RL), in which an agent learns and adapts its strategy by using the feedback of its actions that have been used in the past, has received much attention. Specifically, this theory learns an optimal strategy by repeatedly interacting with the environment.…”

“…In other works, the Q learning‐based algorithm was proposed to address jamming and anti‐jamming strategies against adaptive opponents in multichannel scenarios. In other works, multiarmed bandit–based algorithms, such as the jamming bandit, the dual RL‐based jamming decision, and the positive RL‐orthogonal decomposition were designed by Amuru et al and Zhuansun et al, who aimed to learn an optimal jamming strategy within a few iterations. In jamming applications, the jammer needs to learn and adapt its strategy against its opponent in a timely manner.…”