A Transfer Learning Approach to Discover IDS Configurations Using Deep Neural Networks
Abdulmonem Alshahrani, John A. Clark · 2022
Configuring Intrusion Detection Systems (IDSs) involves balancing functional criteria (e.g., detection rate) and non-functional criteria (e.g., cost). The trade-offs become especially apparent when the network is resource-constrained, as is common for Internet of Things (IoT) networks. Optimisation can be highly computationally intensive. Here we show how transfer learning, which harnesses the experience of previously trained neural networks, can be used to develop a deep learning based proxy model for evaluating candidate IDS configurations more cheaply and accurately.