ISG Digital Case Study Awards™ recognize Persistent Systems

Challenges

As cyber-attacks continue to evolve, the intrusion detection systems (IDS) used to detect these attacks need to stay up to date. Often, attackers target multiple organizations, and the intrusion attack data is spread across them. Considering the sensitive nature of network related data due to regulations and privacy limitations, organizations find it challenging to aggregate data for training machine learning models. Consequently, each organization ends up with a machine learning model that does not achieve its maximum potential. To stay ahead of the attackers, enterprises need to use the most advanced technology and work together in a way that allows them to contribute their insights in a secure manner.

Solutions

Persistent helped the client to develop an innovative solution leveraging privacy-preserving Machine Learning (ML) to build a decentralized Federated Learning system that multiple organizations can use to strengthen their defenses. Federated learning is a decentralized ML technique which allows multiple entities to create a common global ML model without directly sharing data. The solution helps keep individual organizational security logs private but can still enable the gathering of insights and sharing between organizations. While maintaining data privacy, the solution helps build a robust anomaly detector that can learn from attack incidents at individual organizations.

Outcomes

Using this Federated Learning solution, Persistent helped client achieve 38% improvement in accuracy of detecting cyber threats as compared to other methods.

The use of Federated Learning greatly improved the ability of client’s product to effectively identify potential cybersecurity threats and continually improve Intrusion Detection Systems using shared data. For the client, the solution provides improved competitive advantage when compared to other similar products in the market. For the end customer, the solution maintains data privacy while better securing the enterprise.

Technologies Used

• Flower Federated Learning framework
• Custom Federated Learning algorithms
• TensorFlow, Keras for Deep learning Auto-Encoder models
• Amazon Web Services for cloud development