WiMi Hologram Cloud Inc., a global Hologram augmented reality technology provider, has announced implementation of its AI and DL (Deep Learning) based blockchain-assisted intrusion detection technology to enhance Cyber-Physical Systems (CPS) security.
CPS can be integrated with Internet of Things, industrial control systems, and industrial Internet. It is the unity of computing and physical processes and is an intelligent system integrating computing, communication, and control. There are many node devices in the current smart manufacturing CPS system. These devices need to communicate with each other to complete the allocation of resources and improve production collaboration efficiency.
WiMi R&D Centre proposes that CPS brings a high degree of industrial informatisation, and its security becomes a critical technology that cannot be ignored. For this reason, WiMi has developed an intrusion detection technology using blockchain in the CPS environment: WIMI-ProBIDCPS technology. This technology uses AI and DL and can design an effective Intrusion Detection System model for the CPS environment, which, combined with blockchain technology, can improve CPS security.
WIMI-ProBIDCPS uses an Adaptive Harmony Search Algorithm-based feature selection technique. A general regression neural network-based model is applied for intrusion detection and classification. Moreover, the detection efficiency of the GRNN technique has been enhanced by using a hyperparametric optimiser-based algorithm, which improves the intrusion detection results. In addition, blockchain technology enhances security in the CPS environment.
Traditional machine-learning techniques have effectively identified data patterns and detected network attacks in IDSs. However, once the distribution of network nodes is quite large, it is impossible to cover massive data sets effectively, and the performance of detecting network attacks is reduced. Techniques incorporating DL have stimulated IDS mechanisms that enable them to handle network attacks on huge data sets of high complexity today.
WIMI-ProBIDCPS is equipped with behavioral analysis based on trust-related IDS. Node reputation is considered by identifying the variance between two behavioral profiles. Each trust assessment data provided is used to calculate its trust value and then reinforces the learned incentive model. WIMI-ProBIDCPS is designed with the opportunity of collaboration between edge devices and hosts (IoT devices) to achieve IDSs minimising transmission loss and power consumption.
This helps overcome the low classification accuracy and long training time of current deep neural network systems and to achieve an appropriate response to intrusions. For host IDS, a combined deep IDS-based deep belief network is created. A mini-batch gradient descent method is also used for network optimisation and training.
Follow us and Comment on Twitter @TheEE_io