Gaussian Weight Black Widow Optimization based Feature Selection and Deep Learning for Botnet Attack Detection in IoT Healthcare Systems
B.Praveena, Akoijam Priya Devi · 2025
The Internet of Things' (IoT) rapid development has radically changed human existence with the advent of among other ideas are smart devices, a smart grid, smart sector, smart healthcare, and smart cities. Even though the IoT has transformed how individuals utilize technology, it has also given cyber attackers new opportunities. More IoT devices than ever before are connected to the network nowadays. IoT botnet attacks are among the harmful attacks that have emerged and rapidly evolved due to the development of IoT devices. The Centralised Deep Learning (CDL) approach has been widely demonstrated with high classification results for identifying network-based botnet attack detection (AD) in massive amounts of data from IoT networks. An ineffective botnet AD process is caused by the availability of duplicate or irrelevant information and the large dimensionality of network datasets. To determine which intrusion detection system (IDS) features are most important, the Gaussian Weight Black Widow Optimisation (GWBWO) algorithm is suggested in this research. The GWBWO algorithm was influenced by the particular mating customs of Black Widow Spiders (BWS). A Gaussian Weight is created for every sample to balance Local Search (LS) and Global Search (GS). The Gaussian function gives points closer to the central data points larger weights, whereas data points farther away are given lower weights. It extends the training duration and greatly increases the model accuracy. Network traffic classification uses the Deep Neural Network (DNN). The DNN architecture employs several levels of abstractions for hierarchy representations. The BoT-IoT and N-BaIoT datasets were provided by Kaggle and the Machine Learning (ML) Archive at the University of California, Irvine (UCI). These are used to replicate scenarios of network attacks. Recall (R), Precision (P), F1-score, and Accuracy (Acc) were the metrics used to assess DL models using the testing data from the IoT datasets.