Deep learning algorithms have achieved remarkable results in a wide range of tasks, including image classification, language translation, speech recognition, and cybersecurity. These algorithms can learn complex patterns and relationships from large amounts of data, making them highly effective for many applications. However, it is important to recognize that models built using deep learning are not fool proof and can be fooled by carefully crafted input samples. This paper presents the results of a study to explore the use of Generative Adversarial Networks (GANs) in cyber security. The results obtained confirm that GANs enable the generation of synthetic malware samples that can be used to mislead a classification model.