Distributed Denial of Service DDoS attack has been critical for many years. Many organizations have suffered from this type of attack, which, in some incidents, resulted in severe damage. Not only does this attack affect the targeted organizations, but it also impacts legitimate users of those organizations. Likely, technological advancement helps detect and prevent such risks to some extent, even though attackers always find new ways to harm their victims. The last decade has witnessed a dramatic increase in machine learning methods development. Some of the methods have been studied extensively in the area of cyber-attack. One of the technological advancements in machine learning is Transformer models. In this paper, two Transformer models are evaluated in forecasting DDoS attacks. The goal is to test these models and determine which is more suitable to prevent future attacks. The two models are Frequency Enhanced Decomposed Transformer (FEDformer) and Patch Time Series Transformer (PatchTST). The results show that PatchTST is better at forecasting DDoS attacks when compared to FEDformer, even though there is still room for improvement.