Core Concepts
The iTRPL framework proposes an intelligent and behavior-based approach to enhance RPL security by integrating trust mechanisms and multi-agent reinforcement learning. It aims to segregate honest and malicious nodes within a DODAG, making autonomous decisions for optimal network performance.
Abstract
The iTRPL framework introduces a novel approach to enhance RPL security by incorporating trust mechanisms and MARL. It addresses insider attacks by monitoring node behaviors, calculating trust scores, and making informed decisions for DODAG management. Through simulations, iTRPL demonstrates effective decision-making capabilities based on trust scores over multiple epochs.
Routing Protocol for Low Power and Lossy Networks (RPL) is widely used in IoT networks but vulnerable to insider attacks. The iTRPL framework proposes an intelligent solution using trust mechanisms and MARL to secure the network from malicious nodes. By monitoring behaviors, updating trust scores, and making autonomous decisions, iTRPL enhances network security and performance.
Trust computation is crucial in identifying misbehaving nodes within a DODAG network. The iTRPL framework utilizes the Inverse Gompertz function to calculate direct trust based on observed misbehaviors. Indirect trust is provisioned through combining trust opinions from neighboring nodes, enabling informed decision-making for DODAG management.
MARL plays a key role in the iTRPL framework by enabling nodes to make optimal decisions based on calculated rewards from perceived node trust. By utilizing a ϵ-Greedy approach, the root node learns to take actions probabilistically, balancing between exploration and exploitation for efficient DODAG management.
Stats
Trust score depletion for honest nodes is significantly less than selfish or malicious nodes.
Failure rates of non-root nodes vary across different simulation environments.
The ϵ-Greedy approach selects optimal actions stochastically over time.
Variation of node trust over episodes shows different patterns for honest, selfish, and malicious nodes.
Return values differ in highly malicious, moderately malicious, and less malicious DODAG environments.