Universidad Internacional Iberoamericana (International Iberoamerican University, UNIB) researcher Dr. Aman Singh is participating in a study that proposes a security architecture that detects cyber-attacks targeting federated learning systems.
Federated learning, defined as the collaborative training of machine learning models using distributed devices without centralizing data, has attracted growing interest in different fields. It has emerged as an innovative technique, however, this distributed approach also poses security challenges, especially, regarding the integrity of the trained models.
As the demand for advanced machine learning increases, scientists and innovators are actively exploring ways to ensure the privacy, integrity and reliability of learning models. In this regard, researchers are focusing on the potential of software-defined networking (SDN) to improve the security of federated learning systems. SDNs provide centralized control over network resources, including security policies, enabling efficient identification and mitigation of threats at the network level. By implementing the integrated approach of SDN and federated learning, it is possible to create secure and robust learning models.
Therefore, this study proposes an architecture that specifically focuses on cyber-attack detection in federated learning systems by using SDN as an enabler. This novel approach can effectively identify and counter these adversarial actions that undermine the integrity of the trained models.
The researcher used various machine learning models, such as Random Forest, Decision Tree, and K-Nearest Neighbor, to validate the effectiveness of their architecture. These models relied on the N-BaIoT dataset for the simulations and achieved outstanding accuracy rates, with Decision Tree achieving 99.8 % accuracy. These results highlight the strength of the proposed architecture to combat potential cyber-attacks.
Integrating SDN into federated learning architectures can provide significant benefits for secure and efficient distributed learning environments. By using a centralized approach to network management and combining it with the collaborative nature of federated learning, robust intrusion detection and resilient model training is possible.
In conclusion, detecting cyber-attacks on federated learning through SDN monitoring and management opens up new possibilities for training accurate and reliable models while ensuring local data privacy. With further research and advancements, this combined approach can significantly contribute to building reliable and efficient machine learning environments.
To learn more about this study, click here.
To read more research, consult the UNIB repository.
The International Iberoamerican University (UNIB) offers the Master's program in Strategic Management with Specialization in Information Technology. A program that provides professionals with the skills and capabilities needed to manage management positions and lead organizational change projects using ICTs or advise companies that want to join the competitiveness of new businesses. Join the business development by studying our master's program and get ready to take your career to the next level!