Thomas Schneider's Avatar

Thomas Schneider

@thoschneider.bsky.social

https://www.thoschneider.de

68 Followers  |  65 Following  |  3 Posts  |  Joined: 19.02.2025
Posts Following

Posts by Thomas Schneider (@thoschneider.bsky.social)

Call for submissions: #TPMPC2026 (Theory & Practice of MPC)

Submit your latest and coolest results by March 2, 2026.

Aarhus, Denmark, May 18–22, 2026.

Monday: MPC security in practice.

Friday: Symposium celebrating Ivan Damgård’s work.

Links in comments.

10.02.2026 08:18 — 👍 6    🔁 6    💬 1    📌 0
Abstract. Energy-efficient edge devices are essential for the widespread deployment of machine learning (ML) services. However, their limited computational capabilities make local model training infeasible. While cloud-based training offers a scalable alternative, it raises serious privacy concerns when sensitive data is outsourced. Homomorphic Encryption (HE) enables computation directly on encrypted data and has emerged as a promising solution to this privacy challenge. Yet, current HE-based training frameworks face several shortcomings: they often lack support for complex models and non-linear functions, struggle to train over multiple epochs, and require cryptographic expertise from end users. We present HE-SecureNet, a novel framework for privacy-preserving model training on encrypted data in a single-client–server setting, using hybrid HE cryptosystems. Unlike prior HE-based solutions, HE-SecureNet supports advanced models such as Convolutional Neural Networks and handles non-linear operations including ReLU, Softmax, and MaxPooling. It introduces a level-aware training strategy that eliminates costly ciphertext level alignment across epochs. Furthermore, HE-SecureNet automatically converts ONNX models into optimized secure C++ training code, enabling seamless integration into privacy-preserving ML pipeline—without requiring cryptographic knowledge. Experimental results demonstrate the efficiency and practicality of our approach. On the Breast Cancer dataset, HE-SecureNet achieves a 5.2× speedup and 33% higher accuracy compared to ConcreteML (Zama) and TenSEAL (OpenMined). On the MNIST dataset, it reduces CNN training latency by 2× relative to Glyph (Lou et al., NeurIPS’20), and cuts communication overhead by up to 66× on MNIST and 42× on CIFAR-10 compared to MPC-based solutions.

Abstract. Energy-efficient edge devices are essential for the widespread deployment of machine learning (ML) services. However, their limited computational capabilities make local model training infeasible. While cloud-based training offers a scalable alternative, it raises serious privacy concerns when sensitive data is outsourced. Homomorphic Encryption (HE) enables computation directly on encrypted data and has emerged as a promising solution to this privacy challenge. Yet, current HE-based training frameworks face several shortcomings: they often lack support for complex models and non-linear functions, struggle to train over multiple epochs, and require cryptographic expertise from end users. We present HE-SecureNet, a novel framework for privacy-preserving model training on encrypted data in a single-client–server setting, using hybrid HE cryptosystems. Unlike prior HE-based solutions, HE-SecureNet supports advanced models such as Convolutional Neural Networks and handles non-linear operations including ReLU, Softmax, and MaxPooling. It introduces a level-aware training strategy that eliminates costly ciphertext level alignment across epochs. Furthermore, HE-SecureNet automatically converts ONNX models into optimized secure C++ training code, enabling seamless integration into privacy-preserving ML pipeline—without requiring cryptographic knowledge. Experimental results demonstrate the efficiency and practicality of our approach. On the Breast Cancer dataset, HE-SecureNet achieves a 5.2× speedup and 33% higher accuracy compared to ConcreteML (Zama) and TenSEAL (OpenMined). On the MNIST dataset, it reduces CNN training latency by 2× relative to Glyph (Lou et al., NeurIPS’20), and cuts communication overhead by up to 66× on MNIST and 42× on CIFAR-10 compared to MPC-based solutions.

Image showing part 2 of abstract.

Image showing part 2 of abstract.

HE-SecureNet: An Efficient and Usable Framework for Model Training via Homomorphic Encryption (Thomas Schneider, Huan-Chih Wang, Hossein Yalame) ia.cr/2025/1591

05.09.2025 11:49 — 👍 1    🔁 1    💬 0    📌 0
Preview
A suite of privacy-enhancing technologies New technologies and protocols help protect user data while allowing the service providers to securely process that data.,,,European...

Our @erc.europa.eu Starting Grant project Privacy-Preserving Services On The Internet (PSOTI) ended successfully. Its main results are summarized in this "Results in Brief" article. Thanks
@encrypto-group.bsky.social, @cs-tudarmstadt.bsky.social, @tuda.bsky.social! cordis.europa.eu/article/id/4...

23.06.2025 13:56 — 👍 1    🔁 1    💬 1    📌 0

In this upcoming #PETS25 paper, we show how to privately and efficiently compute centrality metrics on multi-layer graphs. A motivating application is private analysis of social interactions on social media platforms. Our open-source implementation scales to graphs with 500k nodes and 5M edges.

05.06.2025 07:59 — 👍 1    🔁 0    💬 0    📌 0

In this upcoming #ASIACCS25 paper, we implement an MPC framework in the memory-safe language Rust. Our focus is on scalability & optimizing memory footprint for secret-sharing-based MPC protocols while maintaining low round complexity. #ENCRYPTO_Group @cs-tudarmstadt.bsky.social @tuda.bsky.social

23.05.2025 08:36 — 👍 5    🔁 0    💬 0    📌 0