A Multiclass Mean-Field Game for Thwarting Misinformation Spread in the Internet of Battlefield Things (IoBT)

TL;DR

This paper models misinformation spread in IoBT systems using a multiclass mean-field game, proposing an equilibrium-based scheme that significantly improves information quality and reduces infection rates.

Contribution

It introduces a novel multiclass mean-field game model for IoBT misinformation control and provides an algorithm to compute the equilibrium in large heterogeneous systems.

Findings

Achieves 1.2-fold increase in information quality over baseline

Reduces infected nodes by 99% compared to baseline

Provides convergence conditions for finite and infinite IoBT systems

Abstract

In this paper, the problem of misinformation propagation is studied for an Internet of Battlefield Things (IoBT) system in which an attacker seeks to inject false information in the IoBT nodes in order to compromise the IoBT operations. In the considered model, each IoBT node seeks to counter the misinformation attack by finding the optimal probability of accepting a given information that minimizes its cost at each time instant. The cost is expressed in terms of the quality of information received as well as the infection cost. The problem is formulated as a mean-field game with multiclass agents which is suitable to model a massive heterogeneous IoBT system. For this game, the mean-field equilibrium is characterized, and an algorithm based on the forward backward sweep method is proposed to find the mean-field equilibrium. Then, the finite IoBT case is considered, and the conditions of convergence of the Nash equilibria in the finite case to the mean-field equilibrium are presented. Numerical results show that the proposed scheme can achieve a 1.2-fold increase in the quality of information (QoI) compared to a baseline scheme in which the IoBT nodes are always transmitting. The results also show that the proposed scheme can reduce the proportion of infected nodes by 99% compared to the baseline.