Backdooring Multimodal Learning

Explore the critical topic of backdooring in multimodal learning systems during this 15-minute IEEE conference talk. Gain insights into the vulnerabilities of machine learning models that process multiple types of data simultaneously, such as text, images, and audio. Discover how malicious actors can exploit these systems through backdoor attacks, potentially compromising their integrity and security. Learn about the latest research findings, detection methods, and mitigation strategies to protect multimodal learning models from these sophisticated threats. Enhance your understanding of the challenges facing AI security in an increasingly complex technological landscape.

Backdooring Multimodal Learning