Most of the currently available network security techniques are not able to cope with the dynamic and increasingly complex nature of the attacks on distributed computer systems. An automated and adaptive defensive tool is imperative for computer networks. One of the emerging solutions for Network Security is the Intrusion Detection System (IDS). However, this technology still faces some challenges such as low detection rates, high false alarm rates and requirement of heavy computational power. To overcome these difficulties, this paper proposes an innovative Machine Learning algorithm called Boosted Modified Probabilistic Neural Network (BMPNN) which utilizes semi-parametric learning model and Adaptive boosting techniques to reduce learning bias and generalization variance in difficult classification. In this paper, BMPNN is implemented as a classifier to detect different types of network anomalies in the KDD-99 benchmark. Extensive experimental outcome indicates that the proposed BMPNN outperforms other state-of-the-art learning algorithms in terms of detection accuracy and model robustness at an affordable computational cost.