Many routing protocols have been proposed for sensor networks, but none addresses the security problems. Directed diffusion is one of the routing protocols for sensor networks. In this paper we study possible attacks on the directed diffusion routing protocol and propose the integration of LEAP, a keying mechanism in to the directed diffusion routing protocol. Our main goal is to identify the potential range of attacks in sensor network and minimize the effects of these attacks. We propose the use of three types of keys to help prevent the various attacks on the sensor nodes. Since the size of each type of key is only 8 bytes, a laptop class adversary can actually break the keys used by the sensor nodes. To minimize this type of attack we extend our approach by proposing new neighbor comparison algorithm. The co-operative nature of the sensor network can be used to prevent the various attacks. Since there are multiple paths available to send the data to the base station, we implemented a simple check mechanism at the intermediate nodes to find out the validity of the messages. In this thesis we show how to eliminate an adversary from participating in the communication. The distance information and the RSSI (Received Signal Strength Indicator) value are used to achieve this. In order to prevent the adversary from faking as a source node, we propose the use of data fusion techniques in the sensor nodes. In data fusion technique, not only the source node collects the data, but some of its neighbors also collect the data from the surroundings. The overheads of our approach are increased communication, computation and storage space. Our simulation result show that the memory space requirement is not an issue for a reasonable number of neighbors, but the communication cost is high in our proposed algorithm.