Adversarial Attack

Adversarial attacks aim to deceive machine learning models by subtly altering input data, causing misclassifications or other erroneous outputs. Current research focuses on developing more robust models and detection methods, exploring various attack strategies across different model architectures (including vision transformers, recurrent neural networks, and graph neural networks) and data types (images, text, signals, and tabular data). Understanding and mitigating these attacks is crucial for ensuring the reliability and security of AI systems in diverse applications, from autonomous vehicles to medical diagnosis and cybersecurity.

Papers

May 28, 2024

Towards Unified Robustness Against Both Backdoor and Adversarial Attacks
Zhenxing Niu, Yuyao Sun, Qiguang Miao, Rong Jin, Gang Hua
Adversarial Attack Adversarial Robustness Backdoor Attack Adversarial Defense General Robustness

May 27, 2024

May 26, 2024

Intruding with Words: Towards Understanding Graph Injection Attacks at the Text Level
Runlin Lei, Yuwei Hu, Yuchen Ren, Zhewei Wei
Adversarial Attack Word List Node Embeddings Graph Injection Attack

May 24, 2024

May 23, 2024

Adversarial Attack

Papers

Towards Unified Robustness Against Both Backdoor and Adversarial Attacks

Spectral regularization for adversarially-robust representation learning

Verifying Properties of Binary Neural Networks Using Sparse Polynomial Optimization

OSLO: One-Shot Label-Only Membership Inference Attacks

Adversarial Attacks on Both Face Recognition and Face Anti-spoofing Models

Intruding with Words: Towards Understanding Graph Injection Attacks at the Text Level

Detecting Adversarial Data via Perturbation Forgery

Uncertainty Measurement of Deep Learning System based on the Convex Hull of Training Sets

Revisit, Extend, and Enhance Hessian-Free Influence Functions

Evaluating and Safeguarding the Adversarial Robustness of Retrieval-Based In-Context Learning

Can Implicit Bias Imply Adversarial Robustness?

Efficient Adversarial Training in LLMs with Continuous Attacks

ART: Automatic Red-teaming for Text-to-Image Models to Protect Benign Users

Large Language Model Sentinel: LLM Agent for Adversarial Purification

Adversarial Attacks on Hidden Tasks in Multi-Task Learning

Are You Copying My Prompt? Protecting the Copyright of Vision Prompt for VPaaS via Watermark

Overcoming the Challenges of Batch Normalization in Federated Learning

SLIFER: Investigating Performance and Robustness of Malware Detection Pipelines

Eidos: Efficient, Imperceptible Adversarial 3D Point Clouds

Towards Transferable Attacks Against Vision-LLMs in Autonomous Driving with Typography