Wireless sensor networks (WSNs) consist of sensor nodes. It is a collection of wild number of low cost device constraint sensor nodes that communicates using wireless medium and they are small in size, low battery power and limited processing capability. This restraint of low electricity power of a sensor node and limited energy capability makes the wireless sensor network failure. Thus, data aggregation is very important techniques in wireless sensor networks and it reduces the energy consumption by eliminating redundancy. The increased deployment of ubiquitous WSN networks has exponentially increased the complexity to detect WSN attacks and protect in oppose to them. In this paper, we consult that collision attack in a network that can be easily launched by a hostile node: a hostile node does not follow the medium access control protocol and purpose collisions attack with neighbor communication by sending a short noise packet. Collision attack means the formation of nodes to access the false data. The data collected from singular nodes is aggregated at a base station or host computer. This attack does not consume much more energy of the attackers but can cause a lot of disruption of the network operation. Due to the wireless broadcast nature, it is not insignificant to identify the attacker. This paper specifies detection algorithms for WSN, which detects collision attack based on the packet flow rate to base station node in the wireless network. Simulation results show that the algorithm have low false toleration and false detection rates and small time to detect attacks.