Certified Robustness

Certified robustness in machine learning aims to provide mathematically guaranteed defenses against adversarial attacks, ensuring model predictions remain reliable even under malicious input perturbations. Current research focuses on improving the accuracy and efficiency of certification methods, particularly through randomized smoothing, diffusion models, and novel neural network architectures like input-convex networks and transformers, exploring multi-norm robustness and addressing challenges like the curse of dimensionality and distribution shifts. This field is crucial for deploying machine learning models in safety-critical applications, as certified robustness offers a higher level of trust and reliability compared to purely empirical evaluations.

Papers

October 4, 2022

Rethinking Lipschitz Neural Networks and Certified Robustness: A Boolean Function Perspective
Bohang Zhang, Du Jiang, Di He, Liwei Wang
Certified Robustness Boolean Function Lipschitz Network Lipschitz Learning

September 29, 2022

Generalizability of Adversarial Robustness Under Distribution Shifts
Kumail Alhamoud, Hasan Abed Al Kader Hammoud, Motasem Alfarra, Bernard Ghanem
Native Robustness Adversarial Robustness Distribution Shift Stronger Generalizability Certified Robustness Adversarial Augmentation DNN Robustness

September 8, 2022

Unraveling the Connections between Privacy and Certified Robustness in Federated Learning Against Poisoning Attacks
Chulin Xie, Yunhui Long, Pin-Yu Chen, Qinbin Li, Arash Nourian, Sanmi Koyejo, Bo Li
Differential Privacy Privacy Policy Poisoning Attack New Connection Certified Robustness Inherent Privacy

July 9, 2022

Adversarial Framework with Certified Robustness for Time-Series Domain via Statistical Features
Taha Belkhouja, Janardhan Rao Doppa
Adversarial Attack Adversarial Example Backdoor Attack Adversarial Framework Certified Robustness Statistical Feature

July 5, 2022

UniCR: Universally Approximated Certified Robustness via Randomized Smoothing
Hanbin Hong, Binghui Wang, Yuan Hong
Adversarial Perturbation Randomized Smoothing Robustness Certification Certified Robustness

June 15, 2022

Can pruning improve certified robustness of neural networks?
Zhangheng Li, Tianlong Chen, Linyi Li, Bo Li, Zhangyang Wang
Neural Network Native Robustness Edge Pruning Certified Robustness

June 8, 2022

Toward Certified Robustness Against Real-World Distribution Shifts
Haoze Wu, Teruhiro Tagomori, Alexander Robey, Fengjun Yang, Nikolai Matni, George Pappas, Hamed Hassani, Corina Pasareanu, Clark Barrett
Deep Neural Network Native Robustness Distribution Shift Certified Robustness

May 12, 2022

Smooth-Reduce: Leveraging Patches for Improved Certified Robustness
Ameya Joshi, Minh Pham, Minsu Cho, Leonid Boytsov, Filipe Condessa, J. Zico Kolter, Chinmay Hegde
Randomized Smoothing Video Classification Class Relevant Patch Certified Robustness Smoothing Algorithm

December 1, 2021

Certified Adversarial Defenses Meet Out-of-Distribution Corruptions: Benchmarking Robustness and Simple Baselines
Jiachen Sun, Akshay Mehra, Bhavya Kailkhura, Pin-Yu Chen, Dan Hendrycks, Jihun Hamm, Z. Morley Mao
Native Robustness Robustness Guarantee Certified Robustness Effective Baseline

November 17, 2021

SmoothMix: Training Confidence-calibrated Smoothed Classifiers for Certified Robustness
Jongheon Jeong, Sejun Park, Minkyu Kim, Heung-Chang Lee, Doguk Kim, Jinwoo Shin
Training Data Adversarial Perturbation Certified Robustness Smoothed Classifier \Ell_p$ Robustness

Certified Robustness

Papers

Rethinking Lipschitz Neural Networks and Certified Robustness: A Boolean Function Perspective

Generalizability of Adversarial Robustness Under Distribution Shifts

Unraveling the Connections between Privacy and Certified Robustness in Federated Learning Against Poisoning Attacks

Adversarial Framework with Certified Robustness for Time-Series Domain via Statistical Features

UniCR: Universally Approximated Certified Robustness via Randomized Smoothing

Can pruning improve certified robustness of neural networks?

Toward Certified Robustness Against Real-World Distribution Shifts

Smooth-Reduce: Leveraging Patches for Improved Certified Robustness

Certified Adversarial Defenses Meet Out-of-Distribution Corruptions: Benchmarking Robustness and Simple Baselines

SmoothMix: Training Confidence-calibrated Smoothed Classifiers for Certified Robustness