Performance Evaluation of Features and Clustering Algorithms for Malware
Authors:
- Houtan FaridiHFHoutan Faridi
- Srivathsan SrinivasagopalanSSSrivathsan Srinivasagopalan
- Rakesh VermaRVRakesh Verma
- 1 November 2018
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 13-22
Abstract
Malware has become a huge problem and obfuscation techniques used by attackers means that signature management for malware becomes too time and space consuming. If we can cluster different versions of the same malware together, we can attain many benefits. In this paper, we study the malware clustering problem systematically using several algorithms, distance functions and sets of features. A real-world, ground-truth dataset and multiple metrics are used to evaluate the performance of these suite of approaches. Agglomerative Hierarchical clustering, Spectral clustering, and DBSCAN emerged as the winning algorithms.Keywords
This publication has 18 references indexed in Scilit:
- Automated Classification and Analysis of Internet Malware2007
- Clustering by Passing Messages Between Data PointsScience, 2007
- Identifying almost identical files using context triggered piecewise hashingDigital Investigation, 2006
- The Similarity MetricIEEE Transactions on Information Theory, 2004
- Estimating the Number of Clusters in a Data Set Via the Gap StatisticJournal of the Royal Statistical Society Series B: Statistical Methodology, 2001
- When Is “Nearest Neighbor” Meaningful?1999
- Syntactic clustering of the WebComputer Networks and ISDN Systems, 1997
- Silhouettes: A graphical aid to the interpretation and validation of cluster analysisJournal of Computational and Applied Mathematics, 1987
- Computer Programs for Hierarchical Polythetic Classification ("Similarity Analyses")The Computer Journal, 1966
- An Ordination of the Upland Forest Communities of Southern WisconsinEcological Monographs, 1957