# SPEED: Secure, PrivatE, and Efficient Deep learning

arXiv: https://arxiv.org/abs/2006.09475 · HTML: https://mlsanity.com/paper/2006.09475/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 16 Jun 2020.

## Built on by (most cited first, 6 of 6)

- [Private and Reliable Neural Network Inference](https://mlsanity.com/paper/2210.15614/) (2022)
- [Efficient Decoding Methods for Language Models on Encrypted Data](https://mlsanity.com/paper/2509.08383/) (2025)
- [Approximate Agreement Algorithms for Byzantine Collaborative Learning](https://mlsanity.com/paper/2504.01504/) (2025)
- [Protecting Data from all Parties: Combining FHE and DP in Federated Learning](https://mlsanity.com/paper/2205.04330/) (2022)
- [When approximate design for fast homomorphic computation provides differential privacy guarantees](https://mlsanity.com/paper/2304.02959/) (2023)
- [SABLE: Secure And Byzantine robust LEarning](https://mlsanity.com/paper/2309.05395/) (2023)

## Abstract (arXiv)

We introduce a deep learning framework able to deal with strong privacy constraints. Based on collaborative learning, differential privacy and homomorphic encryption, the proposed approach advances state-of-the-art of private deep learning against a wider range of threats, in particular the honest-but-curious server assumption. We address threats from both the aggregation server, the global model and potentially colluding data holders. Building upon distributed differential privacy and a homomorphic argmax operator, our method is specifically designed to maintain low communication loads and efficiency. The proposed method is supported by carefully crafted theoretical results. We provide differential privacy guarantees from the point of view of any entity having access to the final model, including colluding data holders, as a function of the ratio of data holders who kept their noise secret. This makes our method practical to real-life scenarios where data holders do not trust any third party to process their datasets nor the other data holders. Crucially the computational burden of the approach is maintained reasonable, and, to the best of our knowledge, our framework is the first one to be efficient enough to investigate deep learning applications while addressing such a large scope of threats. To assess the practical usability of our framework, experiments have been carried out on image datasets in a classification context. We present numerical results that show that the learning procedure is both accurate and private.