WMSNs Data Hidden Anomaly Detection Based on OH-KFJLT Bloom Filter

With the rapid development of the Industrial Internet of Things (IIoT), the volume and size of data handled by Wireless Multimedia Sensor Networks(WMSNs) have increased dramatically. Sensors suffer from damage due to continuous high-load usage and wear, leading to anomalies in sensor data collected and recorded. To solve the problem, this paper proposes an anomaly detection algorithm that is based on the Bloom Filter model, combined with the Optimal Hyperplane-based Kronecker Fast Johnson-Lindenstrauss Transform (OH-KFJLT) mapping and Reciprocal Competition Strategy(RCS), called Optimal Hyperplane Kronecker Fast Johnson-Lindenstrauss Transform Bloom Filter Anomaly Detection(OFBFAD). Firstly, the data is hashed by using the OH-KFJLT mapping method based on optimal hyperplane. Then, the RCS is adopted to de-noise the data. Finally, the Bloom Filter is constructed by 0-1 coding. In the simulation experiments conducted on the NAB, SMD, and COCO benchmark datasets, the False Alarm Rate (FAR) of the OFBFAD algorithm is all lower than 5%. The experimental results show that the OFBFAD has a higher Detection Rate (DR) and lower FAR than the current mainstream anomaly detection algorithm, and can be effectively applied to the anomaly detection of WMSNs data.