PQPrime - Post-Quantum Privacy by Homomorphic Encryption

Code- and Lattice-based Post-Quantum Secure Homomorphic Encryption for Privacy in Information Retrieval, Data Analytics and Machine Learning

In the age of rapid data expansion and the everyday use of AI-based tools, but also the increased awareness of data protection and the prospect of large-scale quantum computers, a way of ensuring post-quantum security of sensitive data, while still allowing its processing for querying and learning algorithms, is highly desired.

PQ-PRIME is a research project on behalf of the Cyberagentur and conducted in cooperation with the German Aerospace Center (DLR). It aims to address the challenges of privacy-preserving information retrieval, data analytics and machine learning using cryptographic methods for processing on encrypted data. A powerful tool to enable the processing of securely encrypted data is Fully Homomorphic Encryption, a type of encryption that allows computations on the ciphertexts, but yet cannot be used in practice due to its high complexity.

The focus of PQ-PRIME is to investigate ways of optimizing existing code-based and lattice-based homomorphic encryption schemes by tailoring them explicitly to specific practical use cases, such as federated learning. Additionally, own cryptographic schemes for data-secure processing are being designed.

Research Areas

PQ-PRIME is divided into three research areas, all developing cryptographic schemes enabling privacy-preserving operations on encrypted data.

Private Information Retrieval and Private Set Intersection

Private Information Retrieval (PIR) describes the problem of a user wanting to retrieve a file without revealing the identity of the downloaded file. The related problem of Private Set Intersection (PSI) refers to two users owning private sets whose intersection should be computed without revealing other information. In the context of PQPrime different PIR and PSI schemes are elaborated, the connection of (symmetric) PIR and PSI is investigated, and homomorphic encryption is applied to enable PIR. This includes the development of novel code-based homomorphic encryption schemes.

Privacy-Preserving Data-Analytics

The goal of privacy-preserving data analytics is to enable secure evaluation of mathematical functions on private data. This research area focuses on tailoring homomorphic encryption to classical statistical methods such as Kalman filtering and clustering. The project investigates both lattice-based and code-based homomorphic encryption approaches, in particular exploring the integration of error-correcting codes to enhance the performance and security of existing schemes.

Privacy-Preserving Machine Learning

Privacy-preserving machine learning enables secure training and inference in machine learning while ensuring data privacy. PQ-PRIME aims to achieve this by customizing existing lattice-based homomorphic encryption to specific machine learning functions. Also novel homomorphic encryption schemes based on codes tailored to machine learning functions and code-based trapdoors for learning problems with applications to private inference and training in machine learning are considered.

Project Team

The project is shared between the Department of Computer Engineering, Professorship for Coding and Cryptography, TU Munich, and the Insitute for Communication and Navigation, Department for Sattellite Networks, German Aerospace Center (DLR). Principal Investigators are Hannes Bartz and Antonia Wachter-Zeh.

• Hannes Bartz (DLR)
• Sebastian Bitzer
• Gökberk Erdogan
• Svenja Lage (DLR)
• Emma Munisamy
• Stefan Ritterhoff
• Antonia Wachter-Zeh