The advancement of telecommunication era is rapidly growing, however, telecomrnfraudsters encouraged by the emerging of these new technologies. Interconnectrnbypass fraud is one of the most sever threats to telecom operators. SubscriberrnIdentity Module Box (SIM-box) fraud is one of an interconnect bypass telecomrnfraud type and uses Voice over IP (VoIP) technology. In addition, it’s difï¬cult torndetect such fraud types with Test Call Generation (TCG) and a traditional typesrnof Fraud Management System (FMS). Both TCG and FMS easily bypassed by thernfraudsters, telecom companies impacted by losing billions of dollars.rnIn this study, Sliding Window (SW) aggregation mode is applied to provide a relevantrndataset instance and reduce detection delay to one hour by using supervisedrnMachine Learning (ML) algorithm. Three supervised ML classiï¬er algorithms werernused, namely Random Forest (RF), Artiï¬cial Neural Network (ANN), and SupportrnVector Machine (SVM) with the two validation techniques 10rn-fold cross-validationrnand supplied test. Call Detail Record (CDR) data were collected, relevant attributesrnwere selected and preprocessing such as data cleaning, integrating and aggregatingrntasks were performed.rnThe experimental results depict that RF classiï¬er using cross-validation on SW aggregationrnmode achieves a better classiï¬cation accuracy (96rn.2rn%). ANN is placed onrnsecond with its overall performance accuracy and its detection delay, SVM algorithmrnusing cross-validation exceeds the desired detection delay (49rn,965rnsecond)rnwith poor performance accuracy. RF classiï¬er algorithm using SW aggregationrnmode overcomes the trade-off detection accuracy and detection delay.