Long distance transportation of various fluid commodities like water, oil, natural gas liquids is achieved through a distribution network of pipelines. Many of these pipelines operates unattended in harsh environments. Therefore, pipes are often susceptible to corrosion, leakage, cracking and third party damage leading to economic and resource infrastructure losses. Thus, early detection and prevention of any further losses is very important. Although many pipeline monitoring techniques exist, the majority of them are based on single sensing modality like acoustic, accelerometer, ultrasound, pressure. This makes the existing techniques unreliable, sensitive to noise and costly. This paper describes a methodology to combine accelerometer and acoustic sensors to increase the detection fidelity of pipeline leakages. The sensors are mounted on the pipe wall at multiple locations. Vibrational and acoustic characteristics obtained from these sensors are fused together through wavelet analysis and classified using kernel SVM and Logistic Regression in order to detect small bursts and leaks in the pipe. The simulation results have confirmed the effectiveness of proposed methodology yielding 90% leak detection accuracy.