Robustness Radius

Robustness radius quantifies a model's resilience to input perturbations, a crucial aspect for reliable machine learning, especially in safety-critical applications. Current research heavily focuses on improving certified robustness radius estimations, primarily using randomized smoothing techniques and their variants (e.g., dual randomized smoothing, partition-based methods) to provide provable guarantees against adversarial attacks. These efforts aim to enhance the accuracy and reliability of predictions in various domains, including computer vision and time series classification, by mitigating the impact of noise and adversarial examples. The development of efficient algorithms for computing robustness radii, particularly in high-dimensional spaces, remains a key challenge with significant implications for the trustworthiness and deployment of machine learning models.

Papers

September 20, 2024

Certified Adversarial Robustness via Partition-based Randomized Smoothing
Hossein Goli, Farzan Farnia
Gaussian Noise Randomized Smoothing Robustness Certificate Robustness Radius

September 4, 2024

Boosting Certified Robustness for Time Series Classification with Efficient Self-Ensemble
Chang Dong, Zhengyang Li, Liangwei Zheng, Weitong Chen, Wei Emma Zhang
Native Robustness Adversarial Robustness Time Series Classification Randomized Smoothing Robustness Certificate Robustness Radius Self Ensemble \Ell_p$ Robustness

August 23, 2024

Deep Learning at the Intersection: Certified Robustness as a Tool for 3D Vision
Gabriel Pérez S, Juan C. Pérez, Motasem Alfarra, Jesús Zarzar, Sara Rojas, Bernard Ghanem, Pablo Arbeláez
Deep Learning Native Robustness Intersection Scenario BRIO Tool Randomized Smoothing Signed Distance Function Voxel Grid Robustness Radius

April 26, 2024

Estimating the Robustness Radius for Randomized Smoothing with 100$\times$ Sample Efficiency
Emmanouil Seferis, Stefanos Kollias, Chih-Hong Cheng
Native Robustness Sample Efficiency Randomized Smoothing Robustness Certificate Robustness Radius

April 15, 2024

Mitigating the Curse of Dimensionality for Certified Robustness via Dual Randomized Smoothing
Song Xia, Yi Yu, Xudong Jiang, Henghui Ding
Data Dimensionality Randomized Smoothing Reversal Curse Dual System Certified Robustness Robustness Radius \Ell_p$ Robustness

September 28, 2023

The Lipschitz-Variance-Margin Tradeoff for Enhanced Randomized Smoothing
Blaise Delattre, Alexandre Araujo, Quentin Barthélemy, Alexandre Allauzen
Native Robustness Randomized Smoothing Robust Classifier Smoothed Classifier Lipschitz Loss Robustness Radius

October 20, 2022

Chaos Theory and Adversarial Robustness
Jonathan S. Kent
Adversarial Attack Adversarial Robustness Cyber Attack Chaos Theory Adversarial Prediction Robustness Radius Loss Input Susceptibility Score

October 2, 2022

Robust Empirical Risk Minimization with Tolerance
Robi Bhattacharjee, Max Hopkins, Akash Kumar, Hantao Yu, Kamalika Chaudhuri
Robust Version Empirical Risk Minimization Robust Learning Robust Model Robust Classification Robustness Radius

March 4, 2022

No More Than 6ft Apart: Robust K-Means via Radius Upper Bounds
Ahmed Imtiaz Humayun, Randall Balestriero, Anastasios Kyrillidis, Richard Baraniuk
Data Clustering K Mean Clustering Method Robust Clustering Feature Centroid Robustness Radius

January 11, 2022

Quantifying Robustness to Adversarial Word Substitutions
Yuting Yang, Pei Huang, FeiFei Ma, Juan Cao, Meishan Zhang, Jian Zhang, Jintao Li
Language Model Native Robustness Robustness Radius Robustness Measurement

December 9, 2021

The Fundamental Limits of Interval Arithmetic for Neural Networks
Matthew Mirman, Maximilian Baader, Martin Vechev
Neural Network Robust Deep Fundamental Limit Interval Arithmetic Robustness Radius Interval Analysis Impossibility Theorem