Research On Intrusion Detection Based On Few-shot Learning

With the continuous development of Internet technology and industry,how to ensure the security of network access equipment has become an important issue.The intrusion detection system can distinguish the normal and abnormal behavior in the network connection,and it is an important means to ensure network security.However,current network intrusion methods often use multiple mechanisms to disguise attacks and evade detection,which poses new challenges to intrusion detection systems.Although many supervised and unsupervised learning algorithms from the field of machine learning and pattern recognition have been used to improve the efficiency of intrusion detection systems,they still have some problems.The algorithm of unsupervised learning does not require a large amount of labeled data,but its detection effect is not good.The detection effect of the supervised learning algorithm is better than that of unsupervised learning,but the supervised learning algorithm needs to use a large number of labeled samples to train the classifier,and it takes time and effort to obtain enough training labeled samples.Another problem of supervised learning is that in a real network,the frequency of abnormal network connection behavior will inevitably be much smaller than the normal network connection request,which leads to the scarcity of abnormal category data in the marked samples,so even if the sufficient labeled samples may also result in poor detection of the algorithm model due to the small sample size of the anomaly category.Few-shot learning is a special case of machine learning.It aims to use only a small number of limited samples for learning and achieve a good learning effect.In order to solve the problems of difficulty in obtaining samples and insuf ficient amount of abnormal samples in network detection,the idea of Few-shot learning algorithm is applied to intrusion detection system.In this paper,based on metric Few-shot learning is used to verify the detection effect of the model on the NSL-KDD and UNSW-NB15 intrusion detection data sets.At the same time,in order to improve the detection effect of the final algorithm model,this paper applies the loss function in face recognition to the algorithm model.Experimental results show that compared w ith other algorithm models,the algorithm model based on Few-shot learning uses the least data,but the detection effect exceeds the detection accuracy and detection rate of other algorithms.