<p>The rapid evolution of industrial systems has made smart industrial infrastructures a prime target for sophisticated cyber threats, posing severe risks to operational continuity and security. To address these challenges, this research presents X-SecuerNet, a transparent and precision-tuned security framework for investigating complex cyberattacks in heterogeneous industrial environments. The proposed framework leverages a combination of Convolutional Neural Networks (CNN) and Attention-based Long Short-Term Memory (ALSTM) to perform hierarchical feature extraction and capture sequential dependencies among complex traffic patterns. A softmax classifier is integrated for multi-class attack detection, ensuring robust classification across diverse cyber threat scenarios with redundant and anomalous feature spaces. To enhance interpretability, X-SecureNet incorporates the Explainable Artificial Intelligence (XAI)-oriented SHapley Additive exPlanations (SHAP) mechanism, explaining the contribution of individual features to the decision-making process. Such interpretations provide valuable insights for cybersecurity analysts and industrial stakeholders to understand the performance of the X-SecureNet framework, thereby enhancing transparency and reliability. The system is rigorously evaluated on CIC-IDS2017 and CIC-IoT2023, two widely used benchmark datasets, which comprise a diverse range of attack instances. The performance of the X-SecureNet is further compared with existing attack detection methods, demonstrating superior detection accuracy and interpretability. These results establish X-SecureNet as a reliable and explainable security solution for safeguarding smart industrial infrastructures, promoting operational integrity and resilience against evolving cyberattacks.</p>