Threat Word
"Threat word" research encompasses the vulnerabilities of various AI systems to malicious manipulation, focusing on how adversarial attacks compromise their functionality, safety, and trustworthiness. Current research investigates these threats across diverse AI applications, including autonomous vehicles (using LiDAR and vision-language models), pricing algorithms, federated learning, and large language models (LLMs), employing techniques like adversarial examples, data poisoning, and prompt injection. Understanding and mitigating these vulnerabilities is crucial for ensuring the responsible development and deployment of AI, impacting fields ranging from transportation safety to economic fairness and online security.
Papers
January 4, 2024
December 29, 2023
December 13, 2023
December 12, 2023
December 10, 2023
November 17, 2023
November 5, 2023
October 22, 2023
August 24, 2023
August 17, 2023
July 14, 2023
July 3, 2023
June 19, 2023
June 4, 2023
April 20, 2023
April 14, 2023
February 3, 2023
January 18, 2023
October 19, 2022