A Data Mining Method Using Deep Learning for Anomaly Detection in Cloud Computing Environment

Abstract EN

Aiming at problems such as slow training speed, poor prediction effect, and unstable detection results of traditional anomaly detection algorithms, a data mining method for anomaly detection based on the deep variational dimensionality reduction model and MapReduce (DMAD-DVDMR) in cloud computing environment is proposed.

First of all, the data are preprocessed by a dimensionality reduction model based on deep variational learning and based on ensuring complete data information as much as possible, the dimensionality of the data is reduced, and the computational pressure is reduced.

Secondly, the data set stored on the Hadoop Distributed File System (HDFS) is logically divided into several data blocks, and the data blocks are processed in parallel through the principle of MapReduce, so the k-distance and LOF value of each data point can only be calculated in each block.

Thirdly, based on stochastic gradient descent, the concept of k-neighboring distance is redefined, thus avoiding the situation where there are greater than or equal to k-repeated points and infinite local density in the data set.

Finally, compared with CNN, DeepAnt, and SVM-IDS algorithms, the accuracy of the scheme is increased by 10.3%, 18.0%, and 17.2%, respectively.

The experimental data set verifies the effectiveness and scalability of the proposed DMAD-DVDMR algorithm.