Visual Classification of Malware by Few-shot Learning

The extent of damage by malware has been multiplying. Many techniques are proposed for detecting malware. However, the usual pattern matching method does not work because when the new malware appeared, many variants are created very soon. In order to catch the new malware, we have to detect and classify them from very few samples. In this paper, we propose a machine learning mechanism that can learn from very few samples of the image of the malware.