Intrusion Detection Model Based On Stacked Sparse Autoencoder And LSTM

With the increasing application of artificial intelligence technology in the Internet of things,the Internet of things faces a great threat to its security and reliability because of its distribution,heterogeneity and other characteristics.With the current development trend of Internet of things,security solutions have become the urgent development direction of Internet of things.In order to ensure the security of Internet of things,intrusion detection is a very key technology.In recent years,machine learning and deep learning have made great breakthroughs in computer vision,natural language processing,image recognition,intelligent translation,recommendation system and so on.Based on this,this paper proposes a research method of the combination of deep learning algorithm and Internet of things intrusion detection technology.This paper focuses on the Internet of things intrusion detection model based on deep learning,in order to improve its detection flexibility,accuracy and efficiency.Through the introduction of deep learning technology,the detection ability of various abnormal attacks in the Internet of things network environment is enhanced,the detection accuracy is improved,the false alarm rate is reduced,and the network security is enhanced.Based on the intrusion detection characteristics of the Internet of things,the deep learning algorithm is used for feature extraction to eliminate redundant features to reduce its impact on the result of feature classification.In order to solve the problems of difficult feature extraction and low detection accuracy in intrusion detection,an intrusion detection model,SSAAE-LSTM,which is applied in the environment of Internet of things and integrates Stacked Sparse Autoencoder(SSAE),attention mechanism and long short term memory network(LSTM),is proposed.The main research contents are as follows:(1)The Sparse Autoencoder based on unsupervised learning is used to compress the input raw data into a low dimensional data representation to remove redundant features and enhance the recognition and extraction of features by intrusion detection system.Stack several Sparse Autoencoder so that they can automatically obtain and learn key features.(2)Using the attention mechanism to reduce the noise of data again and capture important features more accurately can effectively deal with a large number of highdimensional Internet of things intrusion detection samples,accelerate the learning speed of the algorithm and improve the accuracy of the algorithm.(3)Using LSTM for feature classification,the trained SSAAE can also initialize the weights of the potential layer of LSTM,which avoids the random initialization weights in the traditional LSTM method,and further improves the recognition accuracy of LSTM.The effectiveness of the intrusion detection model is verified by simulation experiments.The proposed method can effectively extract the most important features,and can identify and classify them efficiently.