[AAAI 2024] LogFormer: A Pre-train and Tuning Pipeline for Log Anomaly Detection

DGX1

High performance tool for mass log files processing

No description available

No description available

Developed a federated log analysis system using LogFormer with LoRA, applying decentralized training on real-world datasets (HDFS, BGL, OpenStack) to ensure data privacy. Simulated multi-client environments for distributed anomaly detection, achieving high accuracy via secure aggregation and efficient communication.