The industrial Internet of things rapidly through enabling automation, real-time monitoring and data-driven decision-making, has reshaped several industries. Since IIoT devices are networked, they can be attacked by sophisticated malware attacks which renders cybersecurity a significant issue. Existing security methods like rule-based schemes and signature-based IDS are lacking concerning performance against zero-day attacks, are prone to generating high false positives and come at heavy computing expense. This work proposes a Hybrid CNN-GRU Network with Edge Computing to counter such issues to achieve real-time virus detection for IIoT environments. The proposed model enhances threat detection strength without sacrificing computing efficiency by leveraging CNN for spatial feature extraction and GRU for temporal pattern learning. Moreover, hyperparameter tuning is applied based on the Sparrow Search Algorithm for improving classification outcomes. The lower dependence on cloud computing by obtaining low-latency malware detection through the deployment of the model on edge computing hardware. Based on experimental evaluations, the proposed model surpasses traditional classifiers like CNN 93.00 percent, GRU 92.00 percent, LSTM 90.00 percent and SVM 85.00 percent by a significant margin with an accuracy of 99.51 percent. The analysis of the confusion matrix effectively distinguishes between malicious and benign traffic with 100 percent classification accuracy. The superior anomaly detection capabilities of the system are evident from its high precision 99.46 percent, recall 99.56 percent and F1-score 99.51 percent. The edge-based deployment ensures timely response by halving the inference latency from 30 ms to 15 ms based on latency.