Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{208755,
        author = {Chetan S. Chamkure and Prathmesh J. Phuge and Avishkar B. Walke and Jatin S. Karnawat and Arjun A. Khatri},
        title = {AI-Based IoT Intrusion Detection Using Deep Learning},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {no},
        pages = {666-677},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=208755},
        abstract = {Security teams face a growing challenge as IoT devices multiply across smart homes, healthcare systems, industrial plants, and transportation networks, with each new device widening the pool of potential entry points for attackers. Traditional Intrusion Detection Systems (IDS), built on fixed signatures and hand-coded rules, are poorly matched to this setting: IoT traffic is highly varied, devices are resource-limited, data volumes are large, and such systems typically fail to catch zero-day or newly evolving attacks. To close this gap, we present an AI-driven IDS built around a hybrid CNN-LSTM architecture, in which a 1D convolutional layer extracts spatial patterns from individual traffic features while a recurrent LSTM layer captures how those patterns evolve across sequences of packets, jointly aiming for high detection accuracy alongside a low false-alarm rate. The model is built and tested on two widely used IoT security datasets, BoT-IoT and Edge-IIoTset, using a processing pipeline that cleans the raw data, one-hot encodes categorical fields, applies Min-Max scaling, constructs sliding-window sequences, and splits the data into stratified training and test sets. We evaluate performance using accuracy, precision, recall, and F1-score, comparing the hybrid model against Decision Tree, Random Forest, SVM, and a standalone Artificial Neural Network baseline. As detailed in Section 6, the hybrid CNN-LSTM model reaches 99.4% accuracy on BoT-IoT and 98.9% on Edge-IIoTset, with strong per-class F1-scores for DDoS, DoS, and reconnaissance traffic, though performance drops somewhat, while remaining solid, for harder-to-detect information-theft attacks. We close with a discussion of what deploying this system on resource-constrained IoT hardware would require, and point to federated learning and explainable AI as promising directions for future work.},
        keywords = {Internet of Things (IoT); Intrusion Detection System (IDS); Deep Learning; Convolutional Neural Network (CNN); Long Short-Term Memory (LSTM); Network Security; Anomaly Detection; Cybersecurity; Machine Learning.},
        month = {September},
        }