基于卷积神经网络的Android流量分类方法

Classification Method of Android Traffic based on Convolutional Neural Network

深度学习就是机器学习研究的过程,主要通过模拟人脑分析学习的过程对数据进行分析。目前,深度学习技术已经在计算机视觉、语音识别、自然语言处理等领域获得了较大发展,并且随着该技术的不断发展,为网络流量分类和异常检测带来了新的发展方向。移动智能手机与大家的生活息息相关,但是其存在的安全问题也日益凸显。针对传统机器学习算法对于流量分类需要人工提取特征、计算量大的问题,提出了基于卷积神经网络模型的应用程序流量分类算法。首先,将网络流量数据集进行数据预处理,去除无关数据字段,并使数据满足卷积神经网络的输入特性。其次,设计了一种新的卷积神经网络模型,从网络结构、超参数空间以及参数优化方面入手,构造了最优分类模型。该模型通过卷积层自主学习数据特征,解决了传统基于机器学习的流量分类算法中的特征选择问题。最后,通过CICAndmal2017网络公开数据集进行模型测试,相比于传统的机器学习流量分类模型,设计的卷积神经网络模型的查准率和查全率分别提高了2.93%和11.87%,同时在类精度、召回率以及F1分数方面都有较好的提升。

Deep learning is the process of machine learning research,which analyzes data mainly by simulating the process of human brain analysis and learning.At present,deep learning technology has made great progress in the fields of computer vision,speech recognition,natural language processing,etc.,and the continuous development of this technology has brought new development directions for network traffic classification and anomaly detection.Mobile smartphones are closely related to everyone’s life,but their security issues are also increasingly prominent.Aiming at the problem that traditional machine learning algorithms need to manually extract features for traffic classification and have a large amount of calculation,an application traffic classification algorithm based on convolutional neural network model is proposed.Firstly,the data of network traffic is preprocessed to remove the irrelevant data field and make the data meet the input characteristics of convolutional neural network.Then,a new convolution neural network model is designed,and the optimal classification model constructed from the aspects of network structure,superparameter space and parameter optimization.This model solves the problem of feature selection in traditional machine learning-based traffic classification algorithms through autonomous learning of data features by the convolutional layer.Finally,the model test is carried out through CICAndmal2017 network open dataset.Comparison with the traditional machine learning traffic classification model indicates that the designed convolutional neural network model is increased by 2.93%and 11.87%in accuracy and recall respectively.In addition,the class accuracy,recall rate and F1 score are fairly improved.