Deep Learning-Based Privacy Protection for Intrusion Detection in Industrial IoT Networks

The IIoT, characterized by increasing connection and automation in products, services, and processes, has indeed rearranged the type of process that is used in the manufacturing sector. On the other hand, this has also led to concerns about security, as such networks can be hacked, for example, through unlawful accessing of secure information and leakage. It is therefore the aim of the paper to develop a privacy protection paradigm for intrusion detection in IIoT networks employing deep learning techniques. Since the framework is intended for the identification of abnormal activity and intrusion detection in real-time the action employs highly complex deep-learning neural nets. To provide efficient big data processing from IIoT networks, while maintaining a high detection accuracy, a CNN-RNN hybrid deep learning model is used. Furthermore, the model also employs privacy-preserving techniques, for example, differential privacy to protect data during that process. Iodine from extensive tests and trials proves that the system can successfully identify a large number of attacks thus increasing accuracy and reducing the percentage of false positives as compared to other conventional approaches.