Adversarial Pair

Adversarial pairs research focuses on creating and utilizing pairs of data points—one genuine and one subtly altered—to probe the vulnerabilities and robustness of machine learning models. Current research explores this concept across diverse applications, including image generation (using generative adversarial networks or GANs and UNet architectures), speech recognition (leveraging GANs for data augmentation in code-switched languages), and natural language processing (employing techniques like round-trip translation). This work is significant because it reveals weaknesses in existing models, leading to improved model security and reliability, and also enables innovative data augmentation strategies to enhance model performance in data-scarce scenarios.

Papers

September 6, 2023

My Art My Choice: Adversarial Protection Against Unruly AI
Anthony Rhodes, Ram Bhagat, Umur Aybars Ciftci, Ilke Demir
Artificial Intelligence Generative AI Image Generation Art Specific Information Adversarial DEfense Value Laden Choice Large Generative Model Adversarial Clothes Adversarial Pair

May 30, 2022

Adversarial synthesis based data-augmentation for code-switched spoken language identification
Parth Shastri, Chirag Patil, Poorval Wanere, Dr. Shrinivas Mahajan, Dr. Abhishek Bhatt, Dr. Hardik Sailor
Data Augmentation Automatic Speech Recognition Language Identification Speech Domain Speech Classification Adversarial Pair

April 19, 2022

Generating Authentic Adversarial Examples beyond Meaning-preserving with Doubly Round-trip Translation
Siyu Lai, Zhen Yang, Fandong Meng, Xue Zhang, Yufeng Chen, Jinan Xu, Jie Zhou
Adversarial Example Neural Machine Translation Neural Machine Translation Model Round Trip Adversarial Pair

April 8, 2022

Single-level Adversarial Data Synthesis based on Neural Tangent Kernels
Yu-Rong Zhang, Ruei-Yang Su, Sheng Yen Chou, Shan-Hung Wu
Generative Model Generative Adversarial Network Adversarial Learning Neural Tangent Kernel Adversarial Pair

February 7, 2022

Adversarial Attack and Defense for Non-Parametric Two-Sample Tests
Xilie Xu, Jingfeng Zhang, Feng Liu, Masashi Sugiyama, Mohan Kankanhalli
Adversarial Attack Adversarial Vulnerability Adversarial Pair

Adversarial Pair

Papers

My Art My Choice: Adversarial Protection Against Unruly AI

Adversarial synthesis based data-augmentation for code-switched spoken language identification

Generating Authentic Adversarial Examples beyond Meaning-preserving with Doubly Round-trip Translation

Single-level Adversarial Data Synthesis based on Neural Tangent Kernels

Adversarial Attack and Defense for Non-Parametric Two-Sample Tests