Abstract
The Internet of Things (IoT) includes many devices generating vast amounts of data that need extensive computation. IoT has several definitions, but the most popular refers to multiple devices, objects, and sensors all connecting via a network to exchange data. IoT has become more efficient in processing large amounts of data in less time than before because it does not require human intervention. Recently, IoT technologies have improved intelligent systems, such as smart cities, healthcare, smart homes, and more. Unfortunately, IoT faces several security issues and is vulnerable to attacks. To prevent damage or losses, we must detect such anomalies. Internet of Things (IoT) devices are developed daily, leading to increased security vulnerabilities. This work presents an improved deep learning (DL) model for intrusion detection in Internet of Things (IoT) environments to improve accuracy and generalization. It uses convolutional neural network (CNN) capabilities to achieve that. The proposed model was tested on several benchmark datasets and demonstrated notable advances over alternative DL as Long-Short Term Memory (LSTM) and machine learning techniques like Decision Tree (DT). The proposed CNN-based model integrates data augmentation and regularization to prevent overfitting. Furthermore, the model achieves a high precision rate equal to 1, and the average precision to multi-class reaches 82%, which is essential to reduce false positives in real-world applications. This work sets a new standard for future IDS development research and emphasizes how deep learning can be used to improve IoT security. Our enhanced model offers an efficient and scalable way for detecting over 10 attacks to defend IoT networks against constantly changing cyber threats by addressing IoT environments' particular difficulties.