Research map / Federated learning and privacy / Privacy attacks and secure learning
TAPAS: Tricks to Accelerate (encrypted) Prediction As a Service
Amartya Sanyal, Matt Kusner, Adria Gascon, Varun Kanade
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Apply the +1 trick to sparsify encrypted sums (ablation) | TAPAS BNN inference · MNIST | prediction time: 65.1h -> 37.22 h | Improved |
| Apply the +1 trick to sparsify encrypted sums (ablation) | TAPAS BNN inference · Faces | prediction time: 763.5h -> 564h | Improved |
| Apply the +1 trick to sparsify encrypted sums (ablation) | TAPAS BNN inference · Diabetes | prediction time: 283s -> 250s | Improved |
| Apply the +1 trick to sparsify encrypted sums (ablation) | TAPAS BNN inference · Cancer | prediction time: 3.5s -> 3.5s | No clear effect |
| Drop 10% of weights (ablation) | BNN · Cancer test | accuracy: 0.971 -> 0.976 | Improved |
| Drop 10% of weights (ablation) | BNN · Diabetes test | accuracy: 0.549 -> 0.549 | No clear effect |
| Drop 10% of weights (ablation) | BNN · Faces test | accuracy: 0.891 -> 0.879 | Hurt |
| Drop 10% of weights (ablation) | BNN · MNIST test | accuracy: 0.986 -> 0.976 | Hurt |
| Drop 20% of weights (ablation) | BNN · Cancer test | accuracy: 0.971 -> 0.912 | Hurt |
| Drop 20% of weights (ablation) | BNN · Diabetes test | accuracy: 0.549 -> 0.541 | Hurt |
| Drop 20% of weights (ablation) | BNN · Faces test | accuracy: 0.891 -> 0.878 | Hurt |
| Drop 20% of weights (ablation) | BNN · MNIST test | accuracy: 0.986 -> 0.973 | Hurt |
Closest papers by meaning
- United We Stand: Accelerating Privacy-Preserving Neural Inference by Conjunctive Optimization with Interleaved Nexus (AAAI 2024 · similarity 0.83)
- EncryptedLLM: Privacy-Preserving Large Language Model Inference via GPU-Accelerated Fully Homomorphic Encryption (ICML 2025 · similarity 0.82)
- Encryption-Friendly LLM Architecture (ICLR 2025 · similarity 0.82)
- Converting Transformers to Polynomial Form for Secure Inference Over Homomorphic Encryption (ICML 2024 · similarity 0.81)
- FicGCN: Unveiling the Homomorphic Encryption Efficiency from Irregular Graph Convolutional Networks (ICML 2025 · similarity 0.81)
- Secure Distributed Sparse Gaussian Process Models Using Multi-Key Homomorphic Encryption (AAAI 2024 · similarity 0.81)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
