A Denial of Service(DoS) attack is a continuous security risk in cyberspace.They are weaponized with advanced technologies and becomeing more and more powerful as Distributed Denail of Service (DDoS) attacks.DDos is one of the most occuring attack nowadyas and new methods are being needed to be able to detect such attacks.Attackers use many different techniques to perform DDoS attack.Different Dos attack type has different charasterictics and research is still needed to identify such attack.In this paper, we analyse slow DDoS attack types( slowloris and slowhttp attack,etc) and propose a framework to attack them using machine learning techniques.We utilized gain ratio and chi-squared ranking methods to select optimal feature subset for training detection mechanism.CICIDS2017 and CSE-CI-CIDS2018 dataset are used to evaluate the proposed detection mechanism.