Graph Neural Networks For Cyber Threat Analysis: A Self-Adaptive Deep Neural Network Approach Using Gray Bee Optimization

Abstract

In this paper, we propose a novel approach for cyber threat analysis using Graph Neural Networks (GNNs) enhanced by a Self-Adaptive Deep Neural Network (SADNN) and optimized through Gray Bee Optimization (GBO). The primary aim is to identify and mitigate cyber threats effectively by modeling the network as a graph, where nodes represent entities (such as computers, servers, or devices) and edges represent interactions or connections between them. Let G be a connected graph, and let S be a minimum detour dominating degree set of G. A subset T⊆ST is referred to as a forcing subset of S if S is the unique minimum detour dominating degree set containing T. The Self-Adaptive Deep Neural Network dynamically adjusts its parameters based on incoming network traffic, making it highly adaptable to new and evolving threats. GBO, a nature-inspired optimization technique, is used to fine-tune the parameters of the SADNN model, thereby enhancing its performance in detecting complex cyber attacks. The proposed framework not only improves the accuracy of cyber threat detection but also reduces the computational overhead involved in training deep models. We demonstrate the effectiveness of the proposed method on several benchmark datasets and evaluate its performance against state-of-the-art techniques in the field.