Synthesizing Robust Adversarial Examples

Authors: Anish Athalye, Logan Engstrom, Andrew Ilyas, Kevin Kwok

ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We take the first 1000 images in the Image Net validation set, randomly choose a target class for each image, and use EOT to synthesize an adversarial example that is robust over the chosen distribution. We use a fixed λ in our Lagrangian to constrain visual similarity. For each adversarial example, we evaluate over 1000 random transformations sampled from the distribution at evaluation time. Table 1 summarizes the results.
Researcher Affiliation Collaboration 1Massachusetts Institute of Technology 2Lab Six.
Pseudocode No The paper describes algorithms and optimization procedures in text but does not contain structured pseudocode or explicitly labeled algorithm blocks.
Open Source Code No No explicit statement or link to open-source code for the described methodology was found in the paper. The paper only links to a video demonstrating results.
Open Datasets Yes In our experiments, we use Tensor Flow s standard pretrained Inception V3 classifier (Szegedy et al., 2015) which has 78.0% top-1 accuracy on Image Net.
Dataset Splits Yes We take the first 1000 images in the Image Net validation set, randomly choose a target class for each image, and use EOT to synthesize an adversarial example that is robust over the chosen distribution.
Hardware Specification No No specific hardware details (like GPU/CPU models, memory, or cloud instance types) used for running experiments are provided in the paper.
Software Dependencies No The paper mentions 'Tensor Flow s standard pretrained Inception V3 classifier' but does not provide specific version numbers for TensorFlow or any other software dependencies.
Experiment Setup Yes We use a fixed λ in our Lagrangian to constrain visual similarity. ... We searched over several λ values in our Lagrangian for each example / target class pair. In our final evaluation, we used the example with the smallest λ that still maintained 90% adversariality over 100 held out, random transformations. ... we approximate the expectation over transformation by taking the mean loss over batches of size 40; furthermore, due to the computational expense of computing new poses, we reuse up to 80% of the batch at each iteration, but enforce that each batch contain at least 8 new poses.