cPAID Info
cPAID project envisions to develop a cloud-based platform-agnostic defence framework for the holistic protection of AI applications against malicious actions and adversarial attacks. Combining AI-based defence methods, security-and-privacy-by-design, explainable AI, generative AI aims to tackle both poisoning and evasion adversarial attacks. cPAID will design a methodology to :
- Guarantee security-and-privacy-by-design in the AI pipeline
- Thoroughly assess the robustness of ML and DL algorithms against adversarial attacks
- Ensure that EU principles for AI ethics have been considered
- Validate the cPAID components in real-life use case scenarios.
Objectives
Design and development of cPAID platform for the holistic protection and robustness enhancement of AI systems against adversarial attacks.
Definition and implementation of a MLPrivSecOps methodology enhancing the traditional MLOps using Generative Adversarial AI operations and Context Awareness methods for self-improving the system’s robustness.
Research and development of an ML-driven risk management ontology tailored to the needs of AI systems that will provide assurances about the system’s security and robustness.
Design and development an AI-assisted Adversarial Intrusion Detection and Prevention system and advancing information gathering, sharing, and management regarding adversarial attacks.
Investigation and creation of cyber ranges for adversarial attacks to raise awareness, enhance the knowledge of AI professionals and evolve capacity building of EU.
Deployment, validation and evaluation in real-life use-cases.
Communication and dissemination of cPAID results and maximization of its impact.
Partners
