<p>The rapid expansion of the Internet of Things (IoT) adds millions of devices connected to provide intelligent services like remote sensing, actuating and monitoring. However, this widespread connectivity and smart features have also introduced several cybersecurity vulnerabilities. Traditional Machine Learning (ML)-based security solutions are inadequate due to the dynamic nature of the cyber world threats. The new malware types are not detectable by the majority of ML-based security solutions since they rely on static feature extraction. Also, some of the techniques require human expertise and are time-consuming. To address these challenges, this paper proposes a feature-learning-enabled dynamic malware detection and classification method for IoT-centric cybersecurity. The proposed model integrates Convolutional Neural Networks (CNNs) for automated feature extraction and Long Short-Term Memory (LSTM) networks for sequential behaviour analysis. The proposed model ensures adaptive learning and enhances malware detection accuracy. Unlike conventional approaches, our model is not limited to specific malware types. The performance of the proposed approach is evaluated on the MTA-KDD’19 dataset, which provides a realistic representation of legitimate and malicious network traffic. The experimental results demonstrate superior performance, achieving 99.83% precision, 99.90% recall, 99.74% accuracy, and 99.87% F1-score. These findings confirm that the proposed dynamic malware detection framework enhances cyber threat mitigation and ensures a safe IoT system.</p>