Supplementary Material for "What Information Contributes to Log-based Anomaly Detection? Insights from a Configurable Transformer-Based Approach"

This repository contains the supplementary material for the manuscript entitled "What Information Contributes to Log-based Anomaly Detection? Insights from a Configurable Transformer-Based Approach". The material provided is intended to support the findings and methodologies discussed in the paper.

Repository Structure

The repository is organized as follows:

src Folder

• utils: Utility functions and helper scripts. The implementations of positional and temporal encoding methods are included.
• anomaly_bilstm.py: Script for anomaly detection using BiLSTM.
• anomaly_model.py: Defines the anomaly detection model architecture.
• eval_anomaly_bin.py: Evaluation script for binary anomaly detection.
• model.py: Contains the transformer model definitions and configurations.
• sentence_embedding_generation.py: Script for generating sentence embeddings.
• train_anomaly_binary.py: Training script for binary anomaly detection.

Usage

To use the supplementary material, follow these steps:

1. Clone the repository:

   git clone https://github.com/LogAnalyticsResearcher/CfgTransAnomalyDetector.git
   cd CfgTransAnomalyDetector

2. Install the required dependencies: pip install -r requirements.txt

3. Navigate to the src directory and run the desired scripts.

   • First, semantic embeddings for log templates should be generated with sentence_embedding_generation.py.
   • Modify the parameters within train_anomaly_binary.py.
   • Train and test the model: python train_anomaly_binary.py

License

This repository is licensed under the MIT License. See the LICENSE file for more details.

Reference

To be available after the review process.