The research project focuses on developing a Vertical Federated Learning (FL)-based architecture for Cyber Intrusion Detection. This innovative approach leverages federated learning to enhance cybersecurity measures without compromising data privacy
This repository contains the research and all the necessary requirements for the implementation of the architecture. The project explores the use of federated learning techniques to enhance the security and privacy of systems by preserving data locally on devices.
This research project aims to leverage Federated Learning (FL) to enhance cyber intrusion detection capabilities. Traditional cybersecurity systems often face challenges related to data privacy, as centralized models require sharing sensitive data across networks. Federated Learning offers a solution by enabling collaborative model training across distributed devices while keeping data decentralized.
FL allows for model training on data distributed across various verticals (organizations, institutions) while preserving data privacy. This approach not only protects user privacy but also improves the scalability and accuracy of intrusion detection systems. It is particularly crucial in the context of cybersecurity, where data confidentiality is paramount and regulations such as GDPR impose stringent restrictions on data sharing.
Develop an architecture tailored for Intrusion Detection
The primary objective of this research is to design and develop a Vertical Federated Learning (FL) architecture specifically aimed at Cyber Intrusion Detection systems. This involves creating a framework that allows multiple distributed nodes to collaboratively train machine learning models without sharing sensitive data. The architecture will be designed to integrate seamlessly with existing cybersecurity frameworks and systems.
Enhance the privacy and security of the systems
A key objective is to leverage Federated Learning techniques to enhance the privacy and security of intrusion detection systems. By keeping data decentralized and performing model training locally on each node, the architecture aims to mitigate risks associated with data breaches and unauthorized access. Privacy-preserving techniques such as data encryption will be employed to protect sensitive information.
Address the challenges of model and data heterogeneity
Data used for intrusion detection varies significantly across different nodes in terms of format, structure, and quality. One of the objectives is to address these challenges by developing techniques for data augmentation and client seletion in a federated learning setting. This includes adapting machine learning algorithms and optimization strategies to operate efficiently in a distributed, heterogeneous environment.
Demonstrate the feasibility and effectiveness of the architecture
The research aims to empirically validate the proposed Vertical Federated Learning architecture through rigorous evaluations and case studies. This includes benchmarking the performance of the intrusion detection models against centralized approaches, assessing the scalability of the architecture, and evaluating its effectiveness in detecting and mitigating cyber threats. Real-world case studies will be conducted to demonstrate the applicability and benefits of the architecture in practical scenarios.
This research project acknowledges the collaborative supervision and support of two main parties: the Research Center on Scientific and Technical Information (CERIST) and the University of Boumerdes (UMBB). Their expertise and guidance have been instrumental in leveraging the project and efficiently enhancing its prospects. This contribution aims to propose a solution through the collaborative efforts of both parties, ensuring the conception, development, and execution of experiments by providing essential requirements and material resources, including access to a data center, advanced computational infrastructure, and specialized technical expertise. Their support has been crucial in facilitating the successful execution, thus enabling the achievement and conducting significant results in this research project.