MITD-Net: Markov image-based threat detection network.

Area of Science:

• Computer Science
• Cybersecurity
• Artificial Intelligence

Background:

• Malicious user behavior (MUB) poses a significant threat to organizational security, often leading to breaches.
• Existing user activity detection technologies struggle to identify novel or unfamiliar security threats.
• There is a critical need for advanced predictive technologies to counter sophisticated application-based malicious activities.

Purpose of the Study:

• To introduce MITD-Net, a novel method for the effective and efficient prediction of malicious user behavior (MUB).
• To enhance the detection of insider threats by improving upon existing user activity detection methods.
• To develop a proactive system for identifying and mitigating potential security threats within organizations.

Main Methods:

• Developed MITD-Net, a novel method utilizing a MobileNet convolutional neural network (CNN) architecture.
• Extracted features from the CERT r4.2 dataset and converted them into a Markov image for MUB detection.
• Leveraged deep neural networks for computational efficiency and adaptability in low-resource environments.

Main Results:

• MITD-Net demonstrated superior speed and accuracy compared to existing methods for MUB prediction.
• Experimental evaluations on CERT r4.2 datasets confirmed the model's effectiveness in detecting malicious user behavior.
• The proposed approach achieved state-of-the-art or superior performance compared to previous studies.

Conclusions:

• MITD-Net effectively addresses the challenge of predicting harmful user behavior, contributing to proactive threat identification.
• The model enhances overall system security by enabling the early detection and mitigation of potential insider threats.
• The study validates the significance of each component through ablation studies, confirming the robustness of MITD-Net.