O professor Mark Dras, da Universidade Macquarie, na Austrália, irá proferir, na próxima quarta-feira, 27, às 13h30, na sala 2077 do Instituto de Ciências Exatas (ICEx), a palestra “Natural Language Processing: Some Issues in Security and Privacy”. O pesquisador trabalha com machine learning, natural language processing, security, and privacy e está em visita ao Departamento de Ciências da Computação (DCC) da UFMG nesta semana.
Saiba mais sobre a palestra e sobre o professor:
Abstract: There has been a surge of interest in Natural Language Processing in particular since 2022, which saw the release of ChatGPT. Generative AI models are now attracting huge levels of investment and are becoming widely used in industry. However, there are a number of security and privacy issues that language models are susceptible to. In this talk, I will discuss some work with colleagues in these areas. In terms of security, we have looked at the vulnerability of language models (LMs) to backdoor attacks and how model merge can address this; a previously unknown level of vulnerability of LMs to reconstruction attacks, and the limitations of current methods in addressing them; and what characterises adversarial attacks in the language space, and how they differ from the more widely studied image space. In terms of privacy, we have looked at what kinds of inferences of potentially sensitive information can be inferred about the writers of text (authorship identity, native language); and how metric Differential Privacy can help to address these and other privacy issues in NLP.
Bio: Mark Dras is a professor in the School of Computing at Macquarie University, Australia. He works on machine learning and artificial intelligence, with a particular focus on Natural Language Processing, and has more than 150 refereed publications in these areas, in the major venues such as ACL, EMNLP, Computational Linguistics, NeurIPS, IJCAI and ICDM. Mark is currently Treasurer of the Asia-Pacific Chapter of the Association for Computational Linguistics, the premier international scientific and professional society for people working on computational treatment of human language. He has just been appointed to the Australian Research Council’s College of Experts (2025-2027).
