Research On Network Anomaly Detection Based On Collaborative Learning

With the continuous change of information technology,network applications have covered all areas of people's daily life and become an important part of people's daily life,making people's daily life more convenient and efficient.But it also brought a series of problems,such as information leakage and online fraud.Therefore,how to ensure the security of network information is a more concerned issue in current research.And with the rapid development of interconnection technology,network attack methods have gradually changed to intelligence and complexity.Traditional network anomaly detection methods have been unable to meet the requirements of high recognition accuracy and low false alarm rate required by the current environment.The current network anomaly detection technology still has two problems:(1)The traditional machine learning method has the problem of poor model recognition and classification,especially in the field of multi-classification,and requires manual design of feature sets that can accurately reflect traffic characteristics.However,there are still great difficulties in selecting and extracting a feature set that conforms to the model design,and it is also an unresolved problem.(2)The lack of labeled training data,the detection accuracy of the network anomaly detection model trained under the condition of limited label data is low,and how to use the network traffic data of different data sources to jointly train the network anomaly traffic detection model is also an unresolved The problem.Aiming at the above two problems existing in the traditional network anomaly detection model,this paper introduces deep learning technology and collaborative learning technology into the network anomaly detection,optimizes the detection model to achieve a better detection and recognition effect.By studying the application modeling capabilities of convolutional neural networks,BiLSTM networks and collaborative learning technologies in the field of intrusion detection and botnet detection,further optimize the detection model,improve its classification and recognition capabilities,and solve the problems in some traditional network anomaly detection models,And expand the application range of deep learning and collaborative learning technology in the field of network anomaly detection,which has a high research significance.This method effectively solves the problem of manual selection of features and lack of labeled training data.It can use multi-source data to jointly train the same model without sharing private data,which improves the detection performance of the network anomaly detection model and protects the data privacy.