InstaHide: Instance-hiding Schemes for Private Distributed Learning
Authors: Yangsibo Huang, Zhao Song, Kai Li, Sanjeev Arora
ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5. Experiments We have conducted experiments to answer three questions: 1. How much accuracy loss does Insta Hide suffer (Section 5.1)? 2. How is the accuracy loss of Insta Hide compared to differential privacy approaches (Section 5.2)? 3. Can Insta Hide defend against known attacks (Section 5.3)? |
| Researcher Affiliation | Academia | Yangsibo Huang 1 Zhao Song 2 Kai Li 1 Sanjeev Arora 2 1Princeton University 2Princeton University and Institute for Advanced Study. Correspondence to: Yangsibo Huang <yangsibo@princeton.edu>. |
| Pseudocode | Yes | Algorithm 1 Mixup (Zhang et al., 2018) and Algorithm 2 Inside-dataset Insta Hide. |
| Open Source Code | No | The paper states: 'To enable further rigorous study of attacks, we release a challenge dataset of images encrypted using Insta Hide.' and provides a link: 'https://github.com/Hazelsuko07/Insta Hide Challenge'. However, it explicitly states it is releasing a 'challenge dataset' and not the source code for the Insta Hide methodology itself. The prompt specifically requires an 'unambiguous sentence where the authors state they are releasing the code for the work described in this paper'. |
| Open Datasets | Yes | Our main experiments are image classification tasks on four datasets MNIST (Le Cun et al., 2010), CIFAR-10, CIFAR-100 (Krizhevsky, 2009), and Image Net (Deng et al., 2009). |
| Dataset Splits | No | The paper frequently mentions 'test accuracy' and uses terms like 'private training set', but does not explicitly provide details about training/validation/test dataset splits, specific percentages, or sample counts for a validation set. |
| Hardware Specification | No | The paper states 'The implementation uses the Pytorch (Paszke et al., 2019) framework' but does not provide any specific hardware details like GPU or CPU models, memory, or cloud computing specifications used for running the experiments. |
| Software Dependencies | No | The paper mentions 'The implementation uses the Pytorch (Paszke et al., 2019) framework' but does not specify a version number for PyTorch or any other software dependencies, which are required for reproducibility. |
| Experiment Setup | Yes | Hyper-parameters are provided in Appendix F. |