摘要： 提出一种基于稀疏表示的入侵检测算法。将稀疏性约束引入过完备词典学习和编码过程中，使学习得到的稀疏系数可以保持较好的重构性，同时增强判别力。利用判别式K-SVD算法优化过完备词典和线性判别函数，将提取的稀疏特征作为线性分类器的输入，实现入侵检测。实验结果表明，该算法可以获得较低的误报率和较高的检测率，分类性能较好。

关键词: 稀疏编码, 支持向量机, 入侵检测, 奇异值分解, 过完备词典

Abstract: This paper proposes an intrusion detection algorithm based on sparse representation. The sparsity constraints are imposed to over-complete dictionary learning and sparse coding so that the sparse coefficients have better reconstruction and discrimination. The discriminative K-SVD algorithm is exploited to optimize the dictionary and the linear discriminative function, and then extracted features are fed into a linear classifier to implement the intrusion detection. Experimental results show that the algorithm achieves lower false alarm rate and higher detection rate, and it has a good performance in intrusion detection. ?

Key words: sparse coding, Support Vector Machine(SVM), intrusion detection, singular value decomposition, over-complete dictionary

中图分类号: 

• TP309.2