LATIDIA · Robótica

¿Qué tan vulnerable es mi política de aprendizaje? Ataques universales de perturbación adversarial contra las políticas modernas de clonación del comportamiento

arXiv: 2502.03698v5Tipo de anuncio: replace-cross Resumen: El aprendizaje por imitación, también conocido como aprendizaje a partir de demostraciones, es un enfoque popular para entrenar modelos de IA; sin embargo, la vulnerabilidad de estos modelos a la adversaria

La noticia

arXiv:2502.03698v5 Announce Type: replace-cross Abstract: Imitation learning, also known as learning from demonstrations, is a popular approach to train AI models; however, the vulnerability of these models to adversarial attacks remains underexplored. We present the first systematic study of adversarial attacks, across a range of both classic and recently proposed imitation learning algorithms, including Vanilla Behavior Cloning (Vanilla BC), LSTM-GMM, Implicit Behavior Cloning (IBC), Diffusion Policy (DP), and Vector-Quantized Behavior Transformer (VQ-BET). We study the vulnerability of these methods to white-box, grey-box and black-box adversarial perturbations. Our experiments reveal that most existing methods are highly vulnerable to these attacks, including black-box transfer attacks that transfer across algorithms. White-box attacks cause