PANDORA

Welcome to the official repository for Pandora. This document outlines the contents of the uploaded artifacts, the structure of the provided codebases, and instructions for reproduction.

1. Data and Attack Generation Resources

We have provided the fundamental building blocks used in our research to ensure full transparency and reproducibility:

• Raw PCAP Files: The raw packet capture files used for the TTDFIOTIDS2025 dataset analysis have been uploaded.

• Attack Generation Code: The complete source code used to generate attack scenarios is included within the code zip archives.

2. Code Versions

We have uploaded two distinct versions of our intrusion detection system. While the internal loss mechanisms differ, please note that the final detection results are nearly identical across both versions.

Note on Data: For full reproducibility, the TTDFIOTIDS2025 dataset is already included within both code versions. We have ensured that all three datasets are available in both merged and final versions to facilitate immediate testing and verification.

Each version is self-contained and includes its own specific README and Verify Results file for detailed guidance.

Version 1: Manual PMSD Loss

• Description: This implementation uses the PMSD (Prior Multi-Scale Distribution) loss function with manually tuned hyperparameters.

• Use Case: Best for understanding the baseline implementation where parameters are fixed explicitly.

Version 2: Advanced Autobalance Loss (Paper Implementation)

• Description: This is the advanced version introduced in our paper. It utilizes an autobalance loss mechanism to dynamically adjust weights during training, removing the need for manual hyperparameter tuning.

• Use Case: Recommended for leveraging the fully automated approach described in the published methodology.

3. Installation and Execution

Both versions are designed for ease of deployment using Docker. You do not need to manually configure the environment dependencies if you use the provided Dockerfiles.

Steps to Run:

1. Navigate to the directory of the version you wish to test.

2. Build the Docker image using the provided Dockerfile.

3. Run the container.

Specific commands for building and running can be found in the respective README files inside each version's folder.