# On the Robustness of Explanations of Deep Neural Network Models: A Survey

**Authors:** Amlan Jyoti, Karthik Balaji Ganesh, Manoj Gayala, Nandita Lakshmi Tunuguntla, Sandesh Kamath, Vineeth N Balasubramanian  
**arXiv:** [2211.04780](https://arxiv.org/abs/2211.04780)  
**Date:** 2022-11-09  
**Summary page:** https://www.summarizepaper.com/en/paper/2211.04780v1/

## Key points

- Responsible and trustworthy use of machine learning models requires explainability
- Deep Neural Network (DNN) models are increasingly used in risk-sensitive and safety-critical domains
- Many methods have been proposed to explain the decisions made by DNN models
- However, explanations can be distorted or attacked by minor input perturbations
- There has been no effort to assimilate the different methods and metrics proposed to study the robustness of explanations of DNN models
- The paper titled "On the Robustness of Explanations of Deep Neural Network Models: A Survey" presents a comprehensive survey of methods that study, understand, attack, and defend explanations of DNN models
- The paper also provides a detailed review of different metrics used to evaluate explanation methods while describing attributional attack and defense methods
- The authors conclude with lessons and takeaways for the community towards ensuring robust explanations of DNN model predictions

## Summary

The responsible and trustworthy use of machine learning models requires explainability, which has been widely recognized as a cornerstone. With the increasing use of Deep Neural Network (DNN) models in risk-sensitive and safety-critical domains, many methods have been proposed to explain the decisions made by these models. However, recent years have seen concerted efforts that demonstrate how such explanations can be distorted or attacked by minor input perturbations. While there have been many surveys that review explainability methods themselves, there has been no effort hitherto to assimilate the different methods and metrics proposed to study the robustness of explanations of DNN models. In this work titled "On the Robustness of Explanations of Deep Neural Network Models: A Survey," authors Amlan Jyoti, Karthik Balaji Ganesh, Manoj Gayala, Nandita Lakshmi Tunuguntla, Sandesh Kamath, and Vineeth N Balasubramanian present a comprehensive survey of methods that study, understand, attack, and defend explanations of DNN models. The paper also provides a detailed review of different metrics used to evaluate explanation methods while describing attributional attack and defense methods. The authors conclude with lessons and takeaways for the community towards ensuring robust explanations of DNN model predictions. Overall, this work highlights the importance of understanding the robustness of explanations provided by DNN models to ensure their responsible and trustworthy use in various domains.