As the Internet of Things (IoT) grows, ensuring robust security is crucial. Intrusion Detection Systems (IDS) protect IoT networks from various cyber threats. This systematic literature review (SLR) explores the advancements in training and deployment strategies of Artificial Intelligence (AI) based IDS within IoT environments. The study begins by outlining prevalent IoT attacks and developing an updated taxonomy to enhance understanding of these threats. It then examines various IDS architectures and delves into the integration of machine learning (ML) and deep learning (DL) technologies that improve detection capabilities and system responsiveness. Significant emphasis is placed on analyzing different IDS training paradigms-centralized, distributed, and federated learning-and deployment strategies, including in cloud, fog, and edge layers. Their effectiveness within IoT contexts is evaluated comprehensively. Moreover, the review assesses the datasets commonly used for training and validating IDS and discusses key performance metrics and validation measures critical for assessing IDS effectiveness. The study concludes by synthesizing the major challenges facing current IDS implementations in IoT and suggesting future research directions aimed at overcoming these hurdles. This review highlights the technological advancements in IDS and sets the stage for future enhancements in IoT security, emphasizing the integration of innovative training and deployment strategies.
28 March 2025 in WIREs Data Mining and Knowledge Discovery
Industrial Internet of Things (IIoT) deployments demand intrusion-detection systems (IDSs) that are both accurate and lightweight. We propose a two-tier IDS in which a Logistic Regression (LR) model running at the edge classifies the bulk of traffic, while about 16% of low-confidence flows are forwarded to a cloud-hosted 1-D Convolutional Neural Network (CNN) for deeper inspection. A Bowerbird Courtship-Inspired Feature Selection (BBFS) algorithm reduces the input space at both tiers-using a recall-weighted F1 wrapper for LR and a Fisher-score & correlation-penalty filter for the CNN-while a hybrid SMOTE+ENN+LOF strategy mitigates class imbalance. On the WUSTL-IIoT data set, the edge model with balancing plus BBFS attains 99.70% accuracy and 99.71% recall, and the cloud CNN reaches approximately 100.00% accuracy with 99.99% precision and a 99.98% F1-score. Feature reduction lowers LR inference from 52 ms to 38 ms and CNN inference from 27.86 s to 18.28 s on a Raspberry Pi 4 B, demonstrating the framework’s suitability for real-time, resource-constrained IIoT security.
The challenge of feature selection in high-dimensional datasets is critical in enhancing machine learning model performance and interpretability while redu
As a mostly agricultural country, India is deeply concerned about its crop production rate. When crop yields drop, food prices rise, and people who can't afford even potatoes go hungry. Physical examination is a component of conventional approaches for spotting diseases in plants. This approach is exceedingly costly, laborious, and doesn't yield good outcomes. A unique approach for diagnosing potato leaf ailments should be developed to detect sick leaves at the start of their developmental period, which will assist to boost the output and consequently reduce the financial losses suffered by agriculturalists. Fortunately, new Deep Learning (DL) and Machine Learning (ML) models have flourished as a method of reducing plant disease and increasing crop yields, alleviating some of the burdens on farmers. Deep Neural Networks (DNN), a kind of artificial intelligence technology, may be used to identify disease-infected plants so that they can be treated early on, in the bud stage, before the sickness has spread too far. Based on a comprehensive literature assessment, the authors of this research concluded that Convolution Neural Networks (CNN) are superior to other methods for detecting leaf diseases and it is determined that CNN provides the highest achievable accuracy of 91% to 100%.
1 November 2021 in International Journal of Sensors, Wireless Communications and Control
Aim: To design and evaluate the performance of a Grey Wolf Optimization (GWO) based wrapper feature selection applied to the Botnet malware detection system. Background: A botnet is malicious software controlled by a master and used to compromise a distributed set of systems, in turn targeting a victim. Powerful attacks like Distributed Denial of Service (DDoS) can be triggered using a botnet. With the rapid growth of the Internet of Things (IoT) and its omnipresence, the vulnerable IoT devices are also under threat of being a victim or a zombie. Objective: To optimize the listed botnet data traffic features, Grey Wolf Optimization (GWO), in a wrapper model, is used to search the useful features without affecting the classification accuracy. Method: The Botnet dataset consists of a total of 192 Command and Control (C&C) botnet channels HTTP traffic features, and network traffic session-based features. GWO optimization algorithm is used as a wrapper for feature selection, and evaluated on three different classifiers, viz., SVM, KNN, and DT. Results: Decision Tree (DT) and GWO wrapper produced the best results when compared with other classifiers. The output of the research reduces the botnet traffic features to 19 from 192, with an accuracy of 99.73% post the reduction. Conclusion: The proposed DT-GWO wrapper turns out to be an excellent choice for feature reduction for botnet attack detection. The strength of the DT-GWO wrapper is that it is able to retain the near fullfeature accuracy even after a massive reduction of 90.10% of the features.