智能化网络安全防攻击检测中数据抽取和分析

Data extraction and analysis in intelligent network security anti-attack detection

摘要:

针对传统智能化网络安全防攻击检测平台处理数据效率低、误差大等问题,本研究提出一种新型的解决方案。该方案数据抽取模型和大数据分析构建智能化网络安全防攻击检测平台,采用特征模板、卷积神经网络算法模型和条件随机场算法三种方法结合构建出数据抽取模型来抽取网络安全检测数据。其中,利用特征模板提取局部特征向量并进行语句转换得到初始局部向量序列,通过CNN算法对每个网络安全检测数据样本进行卷积和聚合,并提取其特征信息,将语义特征和局部特征相结合经过条件随机场算法进行序列标记,并抽取最优的特征向量序列,最后通过置信传播改进的逻辑回归模型进行分析。实验表明,本研究所提方案克服了现有技术存在的不足,显著提高了处理数据效率和精准度,在数据量为2GB的环境下,经过对数最大似然损失函数得出的损失值只有0.35。

Abstract:

Aiming at the problems of low data processing efficiency and large error of traditional intelligent network security attack detection platform, this paper proposes a new solution. In this scheme, the data extraction model and big data analysis are used to build an intelligent network security attack detection platform. The network security inspection data extraction model is constructed by combining feature template, convolution neural network algorithm model and conditional random field algorithm to extract network security detection data. Among them, the local feature vector is extracted by feature template, and the initial local vector sequence is obtained by sentence conversion. Each network security monitoring data sample is convoluted and aggregated by CNN algorithm, and its feature information is extracted. The semantic feature and local feature are combined, and the sequence is marked by conditional random field algorithm,