Research On DDoS Attack Detection Based On Deep Learning

it is very important to find an effective method for intelligently detecting DDoS attacks.In order to solve this problem,this paper This paper proposes a DDoS attack detection method based on deep learning,and builds a distributed system with Kubernetes and a well-designed neural network model.Compared with traditional machine learning methods,deep learning is highly efficient,malleable,and universal,container-encapsulated deep learning applications can have less computational overhead and better flexibility.The main contents and research results of this paper include:1)Container technology can make full use of various machine resources including CPU memory and disk to achieve efficient use and isolation of resources,while Kubernetes widely recognized and optimistic container orchestration solution in the industry,and natively supports microservice architecture,with good The horizontal capacity expansion and distributed deep learning just needs this horizontal capacity expansion,customize and isolate machine resources,reduce manual operations and provide more efficient training capabilities.In this study we will Use container technology to customize CPU cores and limit memory usage,Kubernetes and TensorFlow.2)Design neural network models and train time series data containing DDoS network attack markers.Research shows that recurrent neural networks(RNN)can effectively detect time series data,while long-and short-term memory networks(LSTM)are such neural networks.The special type can learn long-term time-dependent information and is widely multiple RNN models.The asynchronous stochastic gradient descent algorithm is used to update the weight and deviation parameters.3)Experimental Results Performance Analysis: Utilize the distributed deep learning system we built and the mirrors produced.The experimental parameters can be flexibly adjusted,including custom CPU core and memory maximum usage.Distributed training can shorten the training time of neural network.Through a large number of experiments,the neural network model of CNNLSTM3 has the best performance and output.The degree of fit is best.This paper proposes to the system design from the theory,realizes the recursive neural network model to learn and track the network attack behavior from the time series network traffic characteristics,and uses Kubernetes to build a distributed computing cluster,with the container as the core deployment mode,shortening the training time.This paper proposes a distributed denial of service attack detection method based on cloud computing and deep learning technology.The DDoS detection system implemented in this paper successfully predicts the time series of network traffic.This study is not only academically for the future.The detection of cyber attack behavior provides a reference solution and also solves the actual problem.