Enhancing AI Security with Federated Learning and Advanced Honeypots

Purpose and goal:

This project aims to develop a novel framework for AI security in decentralized learning environments by means of incorporating honeypots into federated learning networks. This will be a starting point in understanding and identifying yet unknown threats and create resilient AI solutions for Swedish organizations.

Expected results and effects: 1. A security framework that incorporates adaptive Honeypots into federated learning networks.

2. An analysis and set of methodologies for assessing the effectiveness and longevity of Honeypots´ deception capabilities within decentralized learning networks. 3. The design of adaptive Honeypots for use in the security framework described above. Approach and implementation: ** Denna text är maskinöversatt **

The project is carried out in 4 work packages (AP). AP1 consists of project management and project advice by senior leaders from the project parties involved. AP2: Develops the framework where adaptive Honeypots will be included as a main component AP3: Develops adaptive Honeypots

AP4: Works with AI Swedens partners to ensure that they are given the opportunity to follow the project and share the results.