The ubiquitous technology known as the Internet of Things is designed to make everyday tasks easier. Wireless sensor based devices with limited power and Internet access became possible with the advent of digital technology and wireless communication. The development of this new scenario heavily relies on Wireless Sensor Networks. These devices are made possible by utilizing the capabilities of sensing, data processing, and node communications. The number of wireless devices with Internet access is growing constantly, and these devices are vulnerable to numerous cyber-attacks due to constrained resource and other reasons. WSNs are vulnerable to many destructive assaults such as flooding, scheduling, and black hole attacks because of their constrained resources. The wireless IDS needs independent research using datasets recorded wireless network test bed in order to detect those threats. Wireless IDS research needs to be done separately from wired IDS research since the majority of the security vulnerability elements of wireless networks are inherent to them and differ from wired networks. In order to effectively track and detect intrusions into those networks, an intelligent, scalable, and fast intrusion detection system are needed due to the complexity and dynamism of the attacks on WSN. Intrusion detection systems (IDS) are a method used in computer security that may identify an intruder in the system and alert the system administrator to any dangerous activity. According to this we have proposed and developed the signature based intelligent IDS for IoT based Wireless sensor network using machine learning. Well-known ensemble machine learning algorithm called XGBoost have been used, for the experiments. A very popular benchmark dataset for IDS in WSN, WSN-DS, have been used for training and testing the ML models. Moreover, employing an XGBoost estimator with all of the preprocessing, feature selection and optimization techniques allowed the achievement of a record breaking high detection accuracy of 99.967% AUC with recursive feature elimination and 99.992% with feature importance feature selection techniques. After the Final model have been trained with Highest AUC features including the optimal hyper parameter selected values, 99.94% of detection accuracy surpassing the scores of any previous relevant studies have been achieved. Keywords: WSN-DS, IDS, GBM, XGBoost, IoT, WSN, ML