Internet is increasingly being used in almost every aspect of our lives and it is becoming a critical resource whose disruption has serious implications. Attacks that aimed at blocking availability of an internet services are generally referred to as denial of service (DoS) attacks. They cause financial losses, as in the case of an attack that prevented users from having steady connectivity to major e-commerce Web sites and imply threat to public safety and national security as in the case of taking down confidential government websites. The consequences of denial of service attacks can be very damaging. Therefore, it is crucial to deter, or otherwise minimize, the damage caused by denial of service attacks. A DoS attack detection system adapted should be able to detect all variants of DoS attacks with efficient computational costs. Thus far various methods have been introduced; however most of them failed to detect new variants of DoS attack and have destitute computation cost. In Multivariate Correlation Analysis based scheme a network features based detection system is introduced. The detection is done based on a normal profile which is generated by applying statistical analysis on the network features. Even though this system has quite good detection rates, some type of DoS attacks are left undetected and have high computation costs. The problems are due to the data used in detection, where the basic features in the original data are in different scales and not optimized. So an efficient detection system is designed by applying Statistical Normalization technique and Particle Swarm Optimization on the raw data to provide efficient detection rates and immaculate computation cost.