Mobile Wireless Sensor Networks (MWSNs) are highly vulnerable to various security threats due to their open communication channels and deployment in unattended or hostile environments. Although clustered MWSNs offer improved energy efficiency, their dynamic and open nature makes them particularly susceptible to attacks. Among these, the Selective Forwarding Attack (SFA) poses a serious threat by selectively dropping packets, thereby disrupting cooperative data transmission and reducing network reliability. This paper proposes a Random Forest (RF)-based SFA detection framework designed for the dynamic nature of clustered MWSNs. The proposed RF-SFAD system integrates Gini Importance (GI) and Recursive Feature Elimination (RFE) for effective feature selection, eliminating irrelevant or redundant data to improve classification performance. It monitors next-hop node behavior using key parameters such as packet loss rate, forwarding rate, packet size, and energy consumption to identify malicious activities indicative of SFA. Experimental evaluations conducted in a simulated MWSN environment demonstrate that the proposed RF-SFAD system achieves a detection accuracy of 98.5%, confirming its effectiveness and robustness in identifying selective forwarding attacks in MWSNs.