Resilience of homomorphically encrypted cloud-computations against integrity attacks: Challenges and countermeasures (REACT)

• Applicants: Professor Dr.-Ing. Tibor Jager; Professor Dr. Moritz Schulze Darup
• Subject Area: Security and Dependability, Operating-, Communication- and Distributed Systems ; Automation, Mechatronics, Control Systems, Intelligent Technical Systems, Robotics
• Term: since 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 503491151

As cloud computing and AI-driven services increasingly rely on outsourced computations, ensuring both the confidentiality and integrity of sensitive data is critical for the resilience of modern systems in industrial control, AI-driven decision-making, smart grids, finance, healthcare, and many other application domains. Fully Homomorphic Encryption (FHE) has emerged as a powerful cryptographic tool, enabling computations on encrypted data without decryption. However, while FHE protects data confidentiality, it remains inherently vulnerable to integrity attacks, in which malicious servers or network adversaries may alter encrypted computations without detection, potentially compromising sensitive or critical applications. The project at hand, "Resilience of Homomorphically Encrypted Cloud Computations Against Integrity Attacks (REACT)", aims to systematically identify and analyze such threats and develop countermeasures to enhance system resilience against attacks. Our research will address two fundamental questions: (1) How can attackers exploit the inherent malleability of FHE to manipulate computations? (2) How can integrity in FHE-based outsourced computations be efficiently ensured without excessive computational overhead? We will investigate novel integrity verification techniques tailored to popular FHE schemes, such as the CKKS scheme, following a new approach that leverages unused SIMD slots and probabilistic integrity checks. Our goal is to provide practical mechanisms to detect and mitigate integrity attacks in privacy-preserving machine learning and networked control applications. Building on our previous work, this follow-up project will enhance system resilience by integrating robust integrity protections into homomorphic encryption frameworks. The results will contribute to more secure and trustworthy applications, supporting their safe deployment in real-world scenarios.

• DFG Programme: Priority Programmes
• Subproject of: SPP 2378: Resilience in Connected Worlds – Mastering Failures, Overload, Attacks, and the Unexpected
• International Connection: South Korea
• Cooperation Partners: Professorin Miran Kim; Dr. Junsoo Kim