The increasing integration of Industrial Internet of Things (IIoT) devices in manufacturing has raised significant cybersecurity concerns, as these systems are particularly vulnerable to sophisticated and evolving attacks. Traditional security methods, such as rule-based and signature-based intrusion detection systems, struggle to detect emerging threats due to their inability to adapt to the complexity and volume of modern network traffic. Standalone Machine Learning (ML) and Deep Learning (DL) models, while useful, fall short in accurately detecting attacks, as feature extraction and classification often require specialized approaches to handle the high-dimensional and multi-source nature of the data. To address these limitations, this research introduces a novel Deep Hybrid Learning (DHL) approach that integrates ML and DL techniques. By combining the strengths of ensemble learning with deep learning architectures, DHL models offer a robust solution for real-time, adaptive cyberattack detection in IIoT environments. We explore three DHL models, two DL models, and two ML models for detecting cyberattacks in IIoT systems, demonstrating that DHL outperforms both ML and DL models in handling the complex, high-dimensional, and multi-source dataset used in this study. The results highlight the innovation and adaptability of DHL in addressing cybersecurity challenges in dynamic and data-rich environments like IIoT.