Nowadays, the Internet of Things (IoT) application is most integrated with our daily lives and society, which are used for exposing the user to threat against their privacy. Moreover, privacy and security of data are some of the major issues in internet-based computing, which become manifolded in IoT for diversified technologies. Furthermore, IoT is adopted in many organizations and academics to protect their assets. So, the current research has proposed a novel machine learning optimized kernel framework for identifying the primary user and malware user using IoT devices. Also, access the primary user and deny malware users using the access history of the IoT device. Additionally, a Support Vector Machine (SVM) is imported to train the database and classify the malware. As well, optimize the kernel function with the help of the Butterfly Optimization Algorithm (BOA). After that, the developed framework analyses the IP address based on the threshold value of normal and malware users, then predicts the primary used based on the stored access history of NU and denies the malware user based on the access history of MU. Finally, IoT devices access the user and the proposed framework is implemented in the python tool. To check the reliability of the proposed framework launch spoofing attacks in the classification layer. Consequently, the performance metrics of the developed technique are compared with other prevailing techniques in terms of detection accuracy, False Prediction Rate (FPR), sensitivity, specificity, precision, and F-measure.