Target Class

Target class manipulation in machine learning models is a significant area of research focusing on how models can be subtly altered to misclassify inputs, often through backdoor attacks or data poisoning. Current research investigates methods to detect and mitigate these attacks, exploring techniques like composite backdoor filtering and generative models for trigger creation, as well as improving model robustness through approaches such as ensembling. Understanding and addressing these vulnerabilities is crucial for ensuring the reliability and security of machine learning systems across various applications, from image recognition to medical diagnosis.

Papers

June 23, 2024

CBPF: Filtering Poisoned Data Based on Composite Backdoor Attack
Hanfeng Xia, Haibo Hong, Ruili Wang
Backdoor Attack Backdoor Poisoning Attack Poisoned Data Composite Attack Target Class

May 9, 2024

Poisoning-based Backdoor Attacks for Arbitrary Target Label with Positive Triggers
Binxiao Huang, Jason Chun Lok, Chang Liu, Ngai Wong
Backdoor Attack Backdoor Defense Adversarial Method Backdoor Poisoning Attack Trigger Design Target Class Trigger Warning

March 22, 2024

Transition Graph Properties of Target Class Classification
Levon Aslanyan, Hasmik Sahakyan
Transition Model Target Class Transition Graph

August 4, 2023

Multi-attacks: Many images $+$ the same adversarial attack $\to$ many target labels
Stanislav Fort
Adversarial Attack Adversarial Example Adversarial Perturbation Step Attack Target Class

November 22, 2022

Pred&Guide: Labeled Target Class Prediction for Guiding Semi-Supervised Domain Adaptation
Megh Manoj Bhalerao, Anurag Singh, Soma Biswas
Domain Adaptation Semi Supervised Unsupervised Domain Adaptation Semi Supervised Domain Adaptation Label Rich Source Target Class

November 21, 2022

Learn to explain yourself, when you can: Equipping Concept Bottleneck Models with the ability to abstain on their concept predictions
Joshua Lockhart, Daniele Magazzeni, Manuela Veloso
Critique Ability Concept Bottleneck Model Semantic Label Concept Based Explanation Concept Prediction Target Class

October 17, 2022

Marksman Backdoor: Backdoor Attacks with Arbitrary Target Class
Khoa D. Doan, Yingjie Lao, Ping Li
Backdoor Attack Backdoor Defense Adversarial Backdoor Target Class

Target Class

Papers

CBPF: Filtering Poisoned Data Based on Composite Backdoor Attack

Poisoning-based Backdoor Attacks for Arbitrary Target Label with Positive Triggers

Transition Graph Properties of Target Class Classification

Multi-attacks: Many images $+$ the same adversarial attack $\to$ many target labels

Pred&Guide: Labeled Target Class Prediction for Guiding Semi-Supervised Domain Adaptation

Learn to explain yourself, when you can: Equipping Concept Bottleneck Models with the ability to abstain on their concept predictions

Marksman Backdoor: Backdoor Attacks with Arbitrary Target Class