Machine-Learning-Based White-Hat Worm Launcher in Botnet Defense System

Abstract

This article proposes a white-hat worm launcher based on machine learning (ML) adaptable to large-scale IoT network for Botnet Defense System (BDS). BDS is a cyber-security system that uses white-hat worms to exterminate malicious botnets. White-hat worms defend an IoT system against malicious bots, the BDS decides the number of white-hat worms, but there is no discussion on the white-hat worms' deployment in IoT network. Therefore, the authors propose a machine-learning-based launcher to launch the white-hat worms effectively along with a divide and conquer algorithm to deploy the launcher to large-scale IoT networks. Then the authors modeled BDS and the launcher with agent-oriented Petri net and confirmed the effect through the simulation of the PN2 model. The result showed that the proposed launcher can reduce the number of infected devices by about 30-40%.

Machine Learning White-Hat Worm Launcher for Tactical Response by Zoning in Botnet Defense System

“…We compared the proposed Launcher with the Launcher of [12]. The notable difference between them having or not having zoning, i.e., the proposed Launcher has zoning, while the Launcher of [12] has no zoning. The simulation results can be seen in Table 3.…”

“…That is, it places worms at random and does not specify concrete nodes. On the other hand, the Launcher of [12] specifies nodes to launch. This Launcher learns how to place white-hat worms based on machine learning.…”

“…However, there is no discussion on the placement of white-hat worms; it should be reasonably considered and configured. Pan et al [12] proposed a machine learning white-hat Worm Launcher for the BDS. It basically has two phases: decision and placement.…”

“…See Figure 6. We used a multitask learning model to implement this tactic in the same way as the second phase of [12].…”

“…The BDS only considers the number of white-hat worms; there is no discussion on the position of white-hat worms. Pan et al [12] proposed a machine-learning (ML)-based white-hat Worm Launcher for the BDS. In this ML-based Launcher, the given network is rigidly divided into several subnetworks with the divide-and-conquer algorithm (D&C algorithm) to adapt the machine learning models, and then, the white-hat worms' positions are predicted by the ML models.…”