step 1: Purpose and Motivation
Why is it important to study?
- Rising incidents of rabies among domestic animals is an increasing concern in Brazilian cities.
- The robustness of neural networks to adversarial examples has received great atttention due to security implications.
Step 2: Problem
What is your research trying to understand or solve?
- The purpose of this study is to investigate the connection between urban population proximity to jungles with wild squirrel populations and the incidence of rabies in domestic animals.
- Despite various attack approaches to crafting visually imperceptible adversarial examples, little has been developed towards a comprehensive measure of robustness.
Step 3: Methods
Discuss only the most significant methods used.
- Using a cross-sectional analysis, this study analyzed the incidence of rabies in Brazilian squirrels from 2007-2015 and measured these results against a similar study conducted in 2011.
- In this paper, we provide a theoretical justification for converting robustness analysis into a local Lipschitz constant estimation problem, and propose to use the Extreme Value Theory for efficient evaluation. Our analysis yields a novel robustness metric called CLEVER, which is short for Cross Lipschitz Extreme Value for nEtwork Robustness. The proposed CLEVER score is attack-agnostic and computationally feasible for large neural networks.
Step 4: Results
State only the most significant results.
- The proximity of a dense urban population to Brazilian squirrel habitats was found to play the greatest role in facilitating the spread of rabies to domestic animals.
- Experimental results on various networks, including ResNet, Inception-v3 and MobileNet, show that (i) CLEVER is aligned with the robustness indication measured by the ℓ 2 \ell_2 ℓ2and ℓ ∞ \ell_\infty ℓ∞ norms of adversarial examples from powerful attacks, and (ii) defended networks using defensive distillation or bounded ReLU indeed achieve better CLEVER scores.
- Quantitative and Qualitative results are presented on several compound expressions, and the experimental demonstrate the feasibility and the potential of EvoGAN.
Step 5: Conclusion
What do the results mean in the context of the problem?
- This study definitively answers the question regarding correlation between proximity to Brazlian squirrels and the transmission of rabies to pets.
- To the best of our knowledge, CLEVER is the first attack-independent robustness metric that can be applied to any neural network classifier.
- The source code is available at https://github/guidao.