The phrase “Industrial Internet of Things” (IIoT) refers to a range of devices, such as actuators, databases, services, industrial applications, and human resources, that are present in the workplace. To maintain the stability of IIoT-connected devices, it is crucial to have a strong security framework in place that safeguards user data and thwarts illicit actions. Therefore, deploying an intrusion detection system is crucial to effectively detect and classify various forms of assaults and anomalous activities inside the IIoT environment. This study employs a deep learning system to forecast several cybersecurity threatens, encompassing distributed denial of service (DDoS), brute force, spoofing, denial of service (DoS), reconnaissance, web-based, and Mirai assaults. The study presents a novel and effective recurrent neural network (RNN) model that precisely forecasts these attack categories by utilizing the Pearson correlation coefficient. To assess the performance of the RNN model, it was compared to various models like the decision tree (DT), random forest (RF), XGBoost, and artificial neural network (ANN). The evaluation was based on accuracy, precision, recall, and F1 score metrics. The efficiency of the existing intrusion detection system (IDS) model in identifying different attacks and anomalies was evaluated using the CICIoT 2023 dataset. To ensure a fair and impartial representation of the data, the synthetic minority over-sampling method (SMOTE) was employed. The test results indicate that the suggested recurrent neural network (RNN) model attained an accuracy of 99.15%. This was accomplished with a learning rate of 0.01, a dropout rate of 0.5, and a prediction time of 30.51 ms. The precision, recall, and F1 scores achieved 99.22%, 99.23%, and 99.23%, respectively. The proposed approach improves the effectiveness of intrusion detection by an average of 4.67% when compared with the most sophisticated machine learning algorithms used for securing Internet of Things (IoT) devices.