面向海量病毒样本家族聚类方法的研究

计算机反病毒厂商每天接收成千上万的病毒样本，如何快速有效地将这些海量样本家族化是一个亟待解决的问题。提出了一种可伸缩性的聚类方法，面对输入海量的病毒样本向量化特征集，使用局部敏感哈希索引技术进行初次快速聚类，使用扩展K均值算法进行二次细致聚类。实验表明该聚类方法在有限牺牲准确度的情况下，大为提高了病毒聚类的时间效率。

Research on familial clustering of massive malware samples

Anti-malware companies receive thousands of malware samples every day, so it becomes more and more pressing to handle these samples timely and effectively. A scalable clustering approach is proposed to group these massive malware samples. LSH algorithm is used to cluster samples rapidly. Extended K-means algorithm is employed to perform accurately clustering. Experimental results show that this approach can improve malware clustering efficiency observably at the cost of little accuracy.