AIDE

The goal of the AIDE project is to define and implement a federated machine learning platform and demonstrate its effectiveness through several case studies. Federated learning allows multiple clients to collaboratively train a machine learning model in a distributed manner. The process consists of the following three steps:

1. Local models are learned on local data and algorithms,
2. Then aggregated into more comprehensive global models that respect the sharing policies of the participating entities,
3. And finally the global models are deployed in the infrastructure of the participating entities.

Federated learning therefore provides an integrated approach for sharing and learning from information without the need to share massive amounts of sensitive data in a centralized server, which provides substantial benefits in terms of security, privacy, and data protection. In addition, federated learning is more efficient in terms of resources as it replaces the sharing of huge datasets with the sharing of model parameters.
In order to create robust federated learning systems, potential vulnerabilities must be carefully evaluated and appropriate risk-based countermeasures must be developed.

A wide range of applications can benefit from federated learning. The AIDE project will start with two cases: one in the field of cybersecurity, where sharing sensitive information is crucial, and a second in Internet of Things (IoT) environments.
For cybersecurity, federated learning by sharing threat information can strengthen penetration testing, improve the synthesis of automatic code repair for vulnerabilities, and optimize malware analysis, threat intelligence, and intrusion detection. For IoT, the focus will be on predictive maintenance.