LMscope / Jbeil

[IEEE SP'24] The Official Implementation of "Jbeil: Temporal Graph-Based Inductive Learning to Infer Lateral Movement in Evolving Enterprise Networks"

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Jbeil: Temporal Graph-Based Inductive Learning to Infer Lateral Movement in Evolving Enterprise Networks

Click Here to Read Full Paper

Accepted and Published at The 2024 IEEE Symposium on Security and Privacy (SP)


Introduction

Jbeil is a data-driven framework to infer Lateral Movement (LM) attacks in evolving enterprise networks. Specifically, Jbeil takes as input time-stamped authentication events (benign events augmented with malicious ones) and output decision on LM activities within the network. The premise of this work is two folds: (i) lies in applying an encoder on a continuous-time evolving graph to produce for each time epoch the embedding of the visible graph nodes; and (ii) a decoder that leverage these embeddings to perform LM link prediction on unseen nodes using an inductive learning technique.

LM

LM

Authors

Joseph Khoury, Đorđe Klisura, Hadi Zanddizari, Gonzalo De La Torre Parra, Peyman Najafirad, Elias Bou-Harb.

Dataset and Preprocessing

Download the public data

Store the csv files in a folder named data/.

  1. Access to the Los Alamos National Laboratory (LANL) Dataset (auth.txt.gz)
  2. Access to the Pivoting Dataset

Preprocess the data

The dense npy format is used to save features in binary format.

For LANL auth.txt.gz dataset:

python utils/preprocess_data.py --data auth

For pivoting dataset:

python utils/preprocess_data.py --data pivoting

Network Graph Map and Graph Features

Check Graph Features Extraction folder.

LM Detection Mechanism

Check Jbeil folder.

Usage:

python train_self_supervised.py --data auth

LM Augmentation Mechanism

Access to the Hopper Lateral Movement Simulator Tool.

Additional resources on the augmentation mechanism will be added soon...

Acknowledgement

Our implementation adapts the code of TGN and Hopper- LM Simulator as the code base and extensively adapts it to our purpose. We thank the authors for sharing their code.

Cite us

@inproceedings{khoury2023jbeil,
title={Jbeil: Temporal Graph-Based Inductive Learning to Infer Lateral Movement in Evolving Enterprise Networks},
author={Khoury, Joseph and Klisura, Dorde and Zanddizari, Hadi and Parra, Gonzalo De La Torre and Najafirad, Peyman and Bou-Harb, Elias},
booktitle={2024 IEEE Symposium on Security and Privacy (SP)},
pages={9--9},
year={2023},
organization={IEEE Computer Society}
}

About

[IEEE SP'24] The Official Implementation of "Jbeil: Temporal Graph-Based Inductive Learning to Infer Lateral Movement in Evolving Enterprise Networks"


Languages

Language:Jupyter Notebook 63.6%Language:Python 36.4%