In this paper Graded Reliance based routing algorithm is proposed to deal with defective nodes in Wireless Sensor Networks (WSN's).The algorithm is intended to validated or build evidence that, by dynamically learning from previous experience and adapting the changes in the operational environment the application performance can be maximized and also enhance operative agility. Quality of service and social network measures are used to evaluate the confidence score of the sensor node. A dynamic modelbased analysis is formulated for best reliance composition, aggregation, and formation to maximize routing performance. The results indicate that reliance based routing approaches yields better performance in terms of message delivery ratio and message delay without incurring substantial message overhead.