Paper Details

Cyber-attacks are keep increasing, and most of these attacks begin with malicious code. Therefore, in orderto reduce the damage caused by cyber-attack, malicious code should be able to be detected and analyzed quickly.According to the AV-test(2017), most of the malicious code found is a variant of existing malware. Therefore, if we canidentify relationship between newly discovered malware and existing malware, the damage caused by cyber-infringementaccidents. According to research, most of the malicious code found is a variant of existing malware. Therefore, if wequickly identify relationship between newly discovered malware and existing malware, then we can reduce the damagecaused by cyber-attacks. In this paper, we propose new method to classify malwares at high-speed. First K-meansalgorithm is used to cluster similar malwares. Then, train a neural network using the clustering result. After training thenetwork, we can classify malwares at high speed. As a result of experiment with 49,561 malwares that actuallydistributed, 84.18% of them are classified into 315 clusters, and the average similarity of clusters was 97.06%. Andaccuracy of classification is over 90%.

Dae-hoon Yoo, Bo-min Choi, Kyung-han Kim, Hong-koo Kang, Jun-hyung Park, “A Study on High-Speed Malware Classification Using Clustering and Neural Network”, International Journal of Advance Engineering and Research Development (IJAERD), Vol. 4, Issue 11, pp. 1290-1295, November 2017.