Agile Proactive Cybercrime Evidence Analysis Model for Digital Forensics

Digital forensics is a critically important area of study dealing with the identification and combating of cyber threats in contemporary networked environments. In this paper, we investigate the possibility of utilizing Large Language Models (LLMs) to examine network traffic categorized as risky according to the University of New South Wales-Network-Based 2015 (UNSW-NB15) dataset. The study employs a multi-phase methodology that combines forensic analysis, evidence extraction, security recommendations, contextual evaluation, and detailed reporting. The results demonstrate high accuracy and qualitative performance across tasks. Automated metrics illustrate the forensic analysis with 95% accuracy, and evidence extraction with 94% precision and 95% coverage. Subjective self-assessment, followed by reviewing 100 examples processed through ChatGPT, shows that outputs have a very high level of clarity (5 out of 5) and relevance (4.5 out of 5). These results highlight the revolutionary role of LLMs in digital forensics with respect to precision, scope, and readability