This paper propose an eXtreme Gradient Boosted Multifactor Ensemble Relevance Vector Node Classification based Attack Detection (XGBMERVNC-AD) Model with the goal of enhancing the performance of attack detection in larger wireless network with lesser communication overhead,. Boosting is a machine learning ensemble meta-algorithm. Boosting is employed in XGBMERVNC-AD Model in order to further increase the stability and accuracy of node classification during the attack identification process. The XGBMERVNC-AD Model considers multifactor such as energy, loss rate and trust value, and cooperativeness of each nodes in network for accurate attack detection. The XGBMERVNC-AD model generates ‘n’ number of weak Relevance Vector Node Classifier (RVNC) results for each an input sensor node based on observed energy, loss rate and trust, cooperativeness value. After that, the weak classifier’s output is combined into strong classifier by considering their error value. This helps for XGBMERVNC-AD model to boost the accuracy of abnormal behavior identification in large wireless IoT network with a lower time usage. Experimental of XGBMERVNC-AD model is accomplished by taking metrics such as attack detection accuracy, communication overhead and packet delivery ratio with respect to various numbers of input sensor nodes.