2026-07-18T08:50:28.924Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:12102 2026-07-12T00:47:01Z 3

Internet of Things Security Framework Based on Light Gradient-Boosting Machine Optimized by Modified Bat Algorithm, Chapter in SIST Smart Innovation, Systems and Technologies: FICTA 2025: Evolution in Computational Intelligence, Springer, volume 477 2026 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/12102 https://link.springer.com/chapter/10.1007/978-3-032-18849-6_3 N. Jovic A. Petrovic L. Jovanovic V. Marevic M. Tomic M. Zivkovic V. Zeljkovic N. Bacanin With the progress of networking technologies and the increasing need for companies to simplify their production process, there are more and more devices that have processors and sensors in them. These devices make up the internet of things (IoT) infrastructure, which is very sensitive to potential malware due to their hardware limitations. This would not be an issue if these networks were not used in scenarios that can jeopardize human life, e.g. the IoT is already being used in healthcare to monitor patients’ conditions. Hence, there is a need to create a robust security system based on artificial intelligence (AI). Therefore, to address this issue, we propose in this manuscript a light gradient boosting machine (LGBM) model, which is optimized by an improved version of the bat algorithm (BA). Combining metaheuristics with machine learning (ML) and deep learning (DL) is an active research area, since the hyperparameters of each ML and DL model should be optimized for a specific challenge. Furthermore, the proposed research also introduces changes in the baseline BA, where the adaptive depletion parameter and the genetic crossover mechanism were adopted. The proposed BA implementation shows admirable performance against the real-world IoT security dataset, but also when compared to high-performing metaheuristic algorithms in terms of standard metrics for classification challenges. bookPart Springer, Cham 30 41 10.1007/978-3-032-18849-6_3 SIST Smart Innovation, Systems and Technologies: FICTA 2025: International Conference on Frontiers of Intelligent Computing: Theory and Applications, volume 477