Research On Intrusion And Malware Classification Technology Based On ELM

Emerging communication technologies,represented by the Intermet,are being deeply integrated in various fields of society and economy rapidly,and are one of the main driving forces for promoting national competitive advantages,social transformation,and consumption upgrading.However,everything has two sides.VWhile people are benefiting from the Internet,the means of cyber attacks are constantly being upgraded and innovated.The methods of cyber attacks are becoming more and more difficult to identify,and the scope of influence is also becoming wider and wider.With the advent of the Intermet age,cyber attacks threaten our normal needs for the network,and may bring about consequences such as infornation leakage and economic losses.As a result,cyber security must face new challenges.Network attacks are generally divided into network and host-side attacks.Therefore,this paper mainly studies network intrusion and malicious code recognition.However,because network intrusion data and malicious code data all have common features,they are large in data volume,attribute redundancy,and linearity between attributes.Related and other characteristics.Therefore,the classification of network intrusion and malicious code is mainly divided into two aspects:data dimension reduction,network intrusion,and malicious code classification.The main work of the thesis includes:1.Research methods for dimension reduction of data include the effects of PCA(feature extraction),rough set and improved rough set algorithm(rough set and improved rough set belong to feature selection)on network intrusion and malicious code classification.2.The effects of extreme learning machine combined with data dimensionality reduction methods(PCA and improved rough set)on network intrusion and malicious code classification and recognition are studied.Because the characteristics of extreme learning machine parameters such as simple setting and good generalization improve the classification speed,and data dimension reduction improves the classification accuracy,PCA and rough set data reduction methods have their own advantages and disadvantages.3.The classification recognition effect of OS-ELM in network intrusion and malicious code recognition is studied.OS-ELM can process datl in batches.After the current data block is trained,it can release the current space,thereby shortening the training time and increasing the efficiency of network intrusion malicious code recognition.OS-ELM combines data dimension reduction methods against network intrusion.With the influence of malicious code classification and recognition accuracy,the method of adding data dimension reduction is higher than the method without data reduction dimension.