mlsanityData to 7 Oct 2026

UFO: Unlocking Ultra-Efficient Quantized Private Inference with Protocol and Algorithm Co-Optimization

arXiv 2602.18758 ↗First posted 21 Feb 2026cs.CR cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Private convolutional neural network (CNN) inference based on secure two-party computation (2PC) suffers from high communication and latency overhead, especially from convolution layers. In this paper, we propose UFO, a quantized 2PC inference framework that jointly optimizes the 2PC protocols and quantization algorithm. UFO features a novel 2PC protocol that systematically combines the efficient Winograd convolution algorithm with quantization to improve inference efficiency. However, we observe that naively combining quantization and Winograd convolution faces the following challenges: 1) From the inference perspective, Winograd transformations introduce extensive additions and require frequent bit width conversions to avoid inference overflow, leading to non-negligible communication overhead; 2) From the training perspective, Winograd transformations introduce weight outliers that make quantization-aware training (QAT) difficult, resulting in inferior model accuracy. To address these challenges, we co-optimize both protocol and algorithm. 1) At the protocol level, we propose a series of graph-level optimizations for 2PC inference to minimize the communication. 2) At the algorithm level, we develop a mixed-precision QAT algorithm based on layer sensitivity to optimize model accuracy given communication constraints. To accommodate the outliers, we further introduce a 2PC-friendly bit re-weighting algorithm to increase the representation range without explicitly increasing bit widths. With extensive experiments, UFO demonstrates 11.7x, 3.6x, and 6.3x communication reduction with 1.29%, 1.16%, and 1.29% higher accuracy compared to state-of-the-art frameworks SiRNN, COINN, and CoPriv, respectively.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 36 of its 66 references that are in the map. First posted on arXiv 21 Feb 2026.

Builds on

36 of 66 references.

References · 36

  1. 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,031
  2. 2018Quantizing deep convolutional networks for efficient inference: A whitepapercited by 507
  3. 2021BRECQ: Pushing the Limit of Post-Training Quantization by Block Reconstructioncited by 464
  4. 2019HAWQ: Hessian AWare Quantization of Neural Networks with Mixed-Precisioncited by 387
  5. 2019HAWQ-V2: Hessian Aware trace-Weighted Quantization of Neural Networkscited by 250
  6. 2024QServe: W4A8KV4 Quantization and System Co-design for Efficient LLM Servingcited by 200
  7. 2018Gazelle: A Low Latency Framework for Secure Neural Network Inferencecited by 193
  8. 2021CrypTen: Secure Multi-Party Computation Meets Machine Learningcited by 131
  9. 2018Mixed Precision Quantization of ConvNets via Differentiable Neural Architecture Searchcited by 122
  10. 2015Fast Algorithms for Convolutional Neural Networkscited by 102
  11. 2020CrypTFlow2: Practical 2-Party Secure Inferencecited by 82
  12. 2022MPCFormer: fast, performant and private Transformer inference with MPCcited by 67
24 more
  1. 2020FracBits: Mixed Precision Quantization via Fractional Bit-Widthscited by 53
  2. 2019XONN: XNOR-based Oblivious Deep Neural Network Inferencecited by 51
  3. 2021DeepReDuce: ReLU Reduction for Fast Private Inferencecited by 50
  4. 2019QUOTIENT: Two-Party Secure Neural Network Training and Predictioncited by 41
  5. 2021BSQ: Exploring Bit-Level Sparsity for Mixed-Precision Neural Network Quantizationcited by 37
  6. 2022Selective Network Linearization for Efficient Private Inferencecited by 31
  7. 2023AutoReP: Automatic ReLU Replacement for Fast Private Network Inferencecited by 29
  8. 2021SIRNN: A Math Library for Secure RNN Inferencecited by 28
  9. 2023Learning to Linearize Deep Neural Networks for Secure and Efficient Private Inferencecited by 20
  10. 2022MPCViT: Searching for Accurate and Efficient MPC-Friendly Vision Transformer with Heterogeneous Attentioncited by 19
  11. 2025LightMamba: Efficient Mamba Acceleration on FPGA with Quantization and Hardware Co-designcited by 11
  12. 2023DeepReShape: Redesigning Neural Networks for Efficient Private Inferencecited by 8
  13. 2024Ditto: Quantization-aware Secure Inference of Transformers upon MPCcited by 7
  14. 2023CoPriv: Network/Protocol Co-Optimization for Communication-Efficient Private Inferencecited by 6
  15. 2023NeuJeans: Private Neural Network Inference with Joint Optimization of Convolution and FHE Bootstrappingcited by 6
  16. 2025Breaking the Layer Barrier: Remodeling Private Transformer Inference with Hybrid CKKS and MPCcited by 6
  17. 2023Falcon: Accelerating Homomorphically Encrypted Convolutions for Efficient Private Mobile Network Inferencecited by 5
  18. 2023PriViT: Vision Transformers for Fast Private Inferencecited by 4
  19. 2020Quantaized Winograd/Toom-Cook Convolution for DNNs: Beyond Canonical Polynomials Basecited by 3
  20. 2022Efficient ML Models for Practical Secure Inferencecited by 2
  21. 2024HEQuant: Marrying Homomorphic Encryption and Quantization for Communication-Efficient Private Inferencecited by 2
  22. 2025CryptPEFT: Efficient and Private Neural Network Inference via Parameter-Efficient Fine-Tuningcited by 2
  23. 2024PrivQuant: Communication-Efficient Private Inference with Quantized Network/Protocol Co-Optimizationcited by 1
  24. 2024Data-Free Group-Wise Fully Quantized Winograd Convolution via Learnable Scalescited by 1

Built on by

Not cited yet.