Mal60: Rare Malware Classification Dataset

The Mal60 dataset consists of 60 malware families, each containing 20 samples, originally collected from a large Korean security company. The dataset reflects the characteristics of targeted malware, which commonly appear in small quantities and often share structural similarities within families. The samples were validated using multiple antivirus engines to increase reliability, including cross-checks with Kaspersky, Bitdefender, and AhnLab, as shown in the paper. Each binary file is transformed into a grayscale image to support image-based malware analysis methodologies such as CNN-based or memory-augmented neural network approaches. The dataset is designed for experiments focused on low-sample malware classification, visualization-based malware analysis, and meta-learning approaches for scarce-data scenarios.