A Novel GPT-Based Framework for Anomaly Detection in System Logs

TL;DR

The paper tackles anomaly detection in large-scale system logs by presenting a GPT-2 based framework that uses structured input derived from Drain event IDs and a Focal Loss objective to handle severe class imbalance. Key contributions include (1) a structured log input design that converts raw logs into Event ID sequences, (2) the integration of Focal Loss to emphasize hard, minority-class samples, and (3) a locally deployed GPT-2 fine-tuning pipeline with robust training/inference procedures. Experimental results show that the optimized GPT-2 system achieves high precision and competitive recall and F1 scores, outperforming the unoptimized model and, in some cases, matching or exceeding GPT-3.5/GPT-4 APIs in offline settings, while offering practical deployment advantages such as reduced latency and data privacy. The work demonstrates a practical pathway for applying pre-trained language models in security contexts by combining structured input representation with loss-function optimization to improve anomaly detection in logs, enabling effective, privacy-preserving, offline deployment with substantial engineering value.

Abstract

Identification of anomalous events within system logs constitutes a pivotal element within the frame- work of cybersecurity defense strategies. However, this process faces numerous challenges, including the management of substantial data volumes, the distribution of anomalies, and the precision of con- ventional methods. To address this issue, the present paper puts forward a proposal for an intelligent detection method for system logs based on Genera- tive Pre-trained Transformers (GPT). The efficacy of this approach is attributable to a combination of structured input design and a Focal Loss op- timization strategy, which collectively result in a substantial enhancement of the performance of log anomaly detection. The initial approach involves the conversion of raw logs into event ID sequences through the use of the Drain parser. Subsequently, the Focal Loss loss function is employed to address the issue of class imbalance. The experimental re- sults demonstrate that the optimized GPT-2 model significantly outperforms the unoptimized model in a range of key metrics, including precision, recall, and F1 score. In specific tasks, comparable or superior performance has been demonstrated to that of the GPT-3.5 API.

Figures (3)

• Figure 1: Architectural Reference Diagram
• Figure 2: Training Loss Curves for the Original and Optimised GPT-2 Models
• Figure 3: Comparative Analysis of Model Optimization Approaches