Automated Malware Classification Using Deep Learning Neural Networks

• Published in: Proceedings (International Conference on Communication Systems and Network Technologies Online) pp. 206 - 212
• Main Authors: Yadav, Surjeet, N, Beemkumar, Loonkar, Shweta
• Format: Conference Proceeding
• Language: English
• Published: IEEE 06.04.2024
• Subjects: Cameras; Data Augmentation; Data Balancing; Malware; Millimeter wave radar; Modular Few-Shot Learning System; Modulation; Radar tracking; Sensors; Signal Modulation Categorization; Traditional Signal Modulation Approaches; Training; Wireless Communication
• ISSN: 2473-5655
• DOI: 10.1109/CSNT60213.2024.10546010

Ingenious applications of millimeter-wave (mmWave) radar for simultaneous monitoring of several targets are possible. It's conceivable, however, that the radar unit itself may malfunction. Improving mmWave radar detection while keeping accurate data classification is difficult. One of the main obstacles to fully capitalizing on mmWave sensing is the time and effort required to construct a model that is compatible with tracking and sensing objectives. Manually annotating mmWave data is time-consuming and often requires subject knowledge since radar frames relate to big events. In this piece, we cover the basics of using a camera for labeling and monitoring in mmWave radar training. A rough sketch of the framework was made during the investigation stage. The effectiveness of the suggested technique is assessed in comparison to other approaches with similar objectives. The conceptual framework has been confirmed by experiments in a range of real-world scenarios. The resulting multi-object tracking system makes use of millimeter wave (mmWave) technology. This technology is unique in that it can tell the difference between a runner, a runner who has stumbled, and a walker. The experiment's results demonstrate that a camera's guided tagging may be used to reliably train a radar model. This trained model continuously exhibits outstanding classification accuracy across a broad variety of examples, not just the ones it was trained on. This paper provides a foundation for further study of hybrid monitoring systems.