In wireless sensor networks, sensing driven nature of data generation and uneven cluster sizes result in unbalanced data traffic load among clusters. Clusters in which the event generation rate is high and/or clusters that have more members than others suffer from congestion and data loss which negatively affect the accuracy of the collected data. In addition, the cluster heads of such clusters exhaust their energy earlier than others, thereby reducing the network lifetime. Hence, the data load among clusters must be balanced to preserve the data accuracy and prolog the network lifetime. This paper presents Secure Load Balancing (SLB) protocol and introduces pseudo-sinks in order to improve data accuracy and bandwidth utilization of wireless sensor networks while still providing secure communication. Simulation results show that, in comparison with traditional cluster based networks, SLB protocol improves the data accuracy and increases the data delivery rate in the presence of security constraints.