Advances, challenges and limitations in the use of automatic anonymization tools for text, audio, images and video

Abstract

The automatic anonymisation of unstructured data, such as images, videos, audio recordings or free text documents, has seen significant advances in recent years, largely due to improvements in Artificial Intelligence models dedicated to the recognition of entities, people and objects. This article begins with a brief introduction to the field of anonymisation in general and, in particular, to the anonymisation of unstructured data, also known as multimodal or multimedia data. Subsequently, a comparative analysis of commercial and open-source tools that perform anonymisation tasks on different types of data is provided. Finally, the advances, challenges and difficulties in the automation of text, speech, voice and image anonymisation, identified in the literature and in the context of a research project carried out by the authors, are presented.