IA GENERATIVA: implicações do processamento de linguagem natural (PLN) e visão computacional na segurança da comunicação GENERATIVE AI: implications of natural language processing (NLP) and computer vision for communication security

Abstract

The advancement of Generative Artificial Intelligence (GAI) has redefined digital communication, especially through models capable of generating synthetic videos (deepfakes) and performing highly realistic voice cloning. These capabilities, which combine progress in Natural Language Processing (NLP) and Computer Vision, also drive the development of tools capable of autonomously creating texts, images, and sounds. Among these tools are automatic translators and virtual assistants, which facilitate interactions and overcome language barriers, while advances in Computer Vision enable the recognition and manipulation of images and videos with a high degree of fidelity. However, these technologies also pose significant risks to information security. Falsified content, such as audio and video deepfakes that accurately imitate the voices and facial expressions of real people, can be used for misinformation and fraud. In addition, large language models can memorize sensitive data during training, increasing the risk of exposure. This article provides a general analysis of the benefits and threats of generative AI in the context of secure communication, based on a critical literature review. It concludes that, despite the gains in efficiency and accessibility, it is urgent to implement mitigation techniques such as encryption and watermarking, as well as ethical regulations, to preserve the integrity, confidentiality, and authenticity of information in the digital environment.