About SHASAI project
The challenge
Artificial intelligence is increasingly deployed in high-risk sectors such as manufacturing, agriculture, healthcare, and transport. While AI improves efficiency and enables advanced capabilities, it also introduces new security risks. AI-based systems face interconnected challenges:
- Hardware vulnerabilities, including side-channel attacks, fault injection, and model extraction
- Software threats, including adversarial attacks, data poisoning, supply chain risks, and privacy threats
- System-level integration gaps, such as regulatory compliance, lifecycle security, and real-time monitoring
Our goal
SHASAI aims to strengthen the security, resilience, and trustworthiness of AI-based systems through a lifecycle-based methodology.
This includes:
- Secure-by-design methods for AI hardware and software
- Tools to assess vulnerabilities across the AI system lifecycle
- AI-based runtime protection for monitoring, detection, and response
- Supply chain security measures
Our tools
By embedding security across the entire lifecycle, SHASAI supports the development of compliant and trustworthy AI systems aligned with the EU AI Act, CRA, NIS2, and related frameworks.
Secure AI supply chain
• Vulnerability analysis tools
• Security analysis tool for data and models
Al for threat detection
• GenAl tools to detect and respond to attacks
Security testing for Al systems
• Blue teaming toolbox
• Red teaming toolbox
Continuous security assessment
• Digital twin platform for security testing
• Automated threat mitigation tools
Our approach
Phase 1: Set-up
• Regulations, standards and ethics for AI (WP1)
• Use case specification, requirement & risk analysis (WP2)
Phase 2: Develop
• Validation and use case scenarios (WP7)
Phase 3: Integrate
Platform for automated secure Al-System lifecycle (WP6)
Phase 4: Validate
• Secure Al supply chain (WP3)
• Security for Al (WP4)
• Al-based security services at operation (WP5)
Demonstration and validation
SHASAI solutions will be validated in three high-risk sectors
These complementary scenarios ensure that SHASAI methods are transferable, scalable, and applicable across critical domains.
Healthcare
Protecting eye-tracking systems used in assistive communication
Agrifood
Securing AI-enabled industrial cutting machines
Automotive
Safeguarding teleoperated last-mile delivery vehicles