SignSGD with Federated Defense: Harnessing Adversarial Attacks through Gradient Sign Decoding
Authors: Chanho Park, Namyoon Lee
ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results demonstrate that sign SGD-FD achieves superior convergence rates compared to traditional algorithms in various adversarial attack scenarios. |
| Researcher Affiliation | Academia | 1Department of Electrical Engineering, POSTECH, Pohang, South Korea 2School of Electrical Engineering, Korea University, Seoul, South Korea. |
| Pseudocode | Yes | Algorithm 1 sign SGD-FD |
| Open Source Code | No | The paper does not provide any concrete access information (specific repository link, explicit code release statement, or code in supplementary materials) for the methodology described. |
| Open Datasets | Yes | The real-world datasets used for image classification simulation are MNIST (Le Cun et al., 1998), CIFAR-10, and CIFAR-100 (Krizhevsky & Hinton, 2009) datasets. |
| Dataset Splits | No | The paper specifies training and test data sizes (e.g., "60,000 images are training data and the remaining 10,000 images are test data" for MNIST) but does not explicitly mention or provide details for a validation split. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. It describes the models (CNN, ResNet-56) but not the hardware used to train them. |
| Software Dependencies | No | The paper mentions using Python and various algorithms, but it does not provide specific software details, such as library names with version numbers (e.g., PyTorch 1.x, TensorFlow 2.x, scikit-learn 0.x). |
| Experiment Setup | Yes | The number of workers M is fixed to 15 and all workers use the same mini-batch size of Bm = 64, m ∈ [M]. The learning rate of each algorithm is carefully selected by comparing the converged test accuracy, where the value is δ = 10−3 and 10−1 for sign SGD-based optimizers and SGD-based optimizers, respectively. To stabilize the learning, we do not actively utilize momentum and weight decay. [...] For ease of implementation, we set the initial period Tin to 50, 100, 500 for the MNIST, CIFAR-10, and CIFAR-100 datasets, respectively. |