The cybersecurity of Machine Learning and Artificial Intelligence solutions: this is the focus of the CybersecH project, an initiative coordinated by Datrix and developed in collaboration with University of Tuscia, co-funded by Cyber 4.0 as part of its core research and innovation projects.
AIA Guard is an end-to-end cybersecurity solution designed to defend against Artificial Intelligence Attacks (AIA) by analyzing the security and privacy risks of the entire workflow of an AI application, from dataset creation to production release of the trained model. Various modules analyze the source code and libraries used for known issues and vulnerabilities; the dataset to check that it does not contain personal and sensitive information; the endpoint where the application is released to make sure it is not an easy target for hackers; and the model to check if it is exposed to data poisoning attacks.
With automated testing and a detailed report with identified risks and vulnerabilities, AIA Guard helps users become aware of issues present and proactively remediate them.
The project is currently in the research and development phase and aims to provide users with a tool that represents the state of the art in preventing attacks against AI systems. A demo of AIA Guard will be made publicly available in the coming months.