Presented by: Ayushi Pandey

Date: Thursday, September 17, 2026
Time: 2:00 pm
Place: Online via Zoom

Speaker:  Ayushi Pandey

Title:  Malware Categorization Using Deep Learning for Multi-Class Classification

Supervisor:  Dr. Issa Traoré

Date:  Thursday, September 17, 2026

Time:  2:00 PM

Location:  Online Via Zoom

Zoom Details:

Meeting link: https://uvic.zoom.us/j/7690949865?pwd=cG5LNkxZYmpXbStvcXVVeERzRE1QQT09&omn=85381174436

Meeting ID: 769 094 9865

Password: 914525

Abstract:  Malware is a significant threat that has grown at an alarming rate especially in recent years. Cyber criminals employ obfuscation techniques to evade security solutions, making it challenging to eliminate threats. Obfuscated malware is adept at hiding itself to evade detection and eradication. Conventional detection methods like heuristic-based or signature-based systems face challenges against such type of malwares, as the traces left in the system are harder to detect and analyze. This project introduces an innovative approach to enhance the detection of obfuscated malware with the integration of deep learning models using multi-class classification. We studied and compared two different algorithms: Multilayer Perceptron (MLP) and 1D Convolutional Neural Network (1D-CNN), using the CIC-MalMem-2022 dataset which is designed to test obfuscated malware detection methods. The dataset consists of spyware, ransomware, and Trojan malware samples. The experimental results demonstrated the effectiveness of 1D-CNN model in achieving high accuracy rates across all evaluation metrics. The 1D-CNN achieved an accuracy of 81.05%, slightly higher than that of MLP (79.43%). Similarly, the CNN also showed an improved precision (81.30% vs. 79.43%), recall (81.05% vs. 79.43%), and F1-score (81.12% vs. 79.43%), demonstrating its stronger ability to learn the spatial feature patterns from the input data for multi-class classification.