Design and modeling of an RL based AI agent for jamming mitigation in communication jamming environment

Project Title

Design and modeling of an RL based AI agent for jamming mitigation in communication jamming environment

Project Area of Specialization Artificial IntelligenceProject Summary

Wireless communication has become increasingly
vulnerable to jamming attacks due to its nonrestrictive sharing
properties. An attacker can interfere with a communication link
resulting in degradation of a network, denial of service, or even
deception. The advent of reactive jammers begs the need for
an intelligent jamming mitigation system. This paper pertains
to the design, modeling, and simulation of an autonomous
Reinforcement learning (RL) based AI agent which can be
employed for anti-jamming applications in a communication
jamming environment. The AI agent would not require jammer
type classification before initiating appropriate anti-jamming
schemes and would autonomously converge to optimum jamming
mitigation scheme using RL. This paper briefly discusses key Q
learning concepts, jamming, anti-jamming strategies, and implementing
reinforcement learning to counter jamming attacks. The
final section focuses on practical implementation using software-defined
radio (SDR)

Project Objectives

The objective of this project is to counter a jamming attack by autonomously detecting an attack, finding an optimal strategy, and implementing it in the most efficient manner. The reinforcement learning agent will learn to counter the attack itself and no prior knowledge of the jamming attack will be required. 

Project Implementation Method

Previously there was a need for an operator to monitor the spectrum and look for any jamming activity that is distorting the communication link. If detected, he would then study the jamming strategy carried out by the jammer followed by implementing countermeasures to escape the attack. The operator was required to constantly monitor the signal spectrum for any anomalies.

Benefits of the Project

There is no need for an operator to constantly look for any attacker activity. The reinforcement learning agent will continuously monitor the communication link. It will also be autonomously carrying out countermeasures without any prior knowledge fed regarding the attacker. The agent will also continuously learn from its actions and interactions with the environment. This will enable the agent to learn better ways to counter an attack most efficiently.

Technical Details of Final Deliverable

The final deliverables will include programming an agent using python. The agent will then be tested in a communication jamming environment. The testing will be carried out in the OpenAI gym. The results will be gathered. Lastly, a real-life demonstration of this agent will be carried out using SDR in an actual jamming environment.

Final Deliverable of the Project HW/SW integrated systemCore Industry TelecommunicationOther IndustriesCore Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Partnerships to achieve the GoalRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	70000
SDR	Equipment	2	35000	70000