‘Curse of dimensionality’ - an unresolved challenge in the design of an intelligent system makes dimensionality reduction a significant topic of research for the identification of informative features from high-dimensional data sets. This paper presents a novel feature selection technique based on Rough Sets (RS) and few interesting properties of Hypergraph (RSHGT), such as minimal transversal and vertex linearity for the identification of the optimal feature subset. Experiments were carried out using KDD cup 1999 intrusion dataset obtained from the UCI repository. Validation using Weka tool shows the dominance of RSHGT over the existing feature selection techniques with respect to the reduct size, classifier accuracy and time complexity. To summarize, RSHGT was found to be flexible, accommodative and computationally attractive for high dimensional data sets.