MalDetect is a deep learning malware detection system built using the EMBER dataset of 1.1 million executables. This neural network was trained on over 600,000 Portable Executable samples and achieved an accuracy of 97.8% in detecting a file as malicious.
Features include a handpicked selection of 100 PE libraries, boolean file properties (has_imports, has_exports, has_tls, etc.), 64 bytes of the PE entry point (used as a signature), and other features relevant to malware detection.
This project is released under the MIT license. Source code provided by EMBER is covered by the GNU Affero General Public License version 3 (AGPL-v3). The data files provided by EMBER are covered by the MIT License.