Abstarct: Wireless sensor networks are one of the applied and attractive technologies that have attracted the attention of researchers in recent years. These networks because of their inherent advantages such as lower cost and easier deployment on the environment, to play a role in a wide range of applications. However, resource constrains, such as limited processing power, memory and energy are main challenge in WSN design and application. Given that WSNs are often used in remote and unprotected locations or even hostile operating conditions, they are highly susceptible to intrusions and security attacks, which, given their limited resources, reduces their performance considerably. Therefore, security in WSNs has become an important issue, and actually security is considered as one of the essential parameters in quality of service (QOS), especially if these networks are involved in critical processes. Many security solutions are designed for WSNs to specific attacks. These security mechanisms are able to guarantee security at some levels; however, they cannot be completely prevented from most of the security attacks. The intrusion detection system (IDS) is one of defense method against attacks, which is actually the second line of defense against invaders after intrusion prevention systems (IPS), and its task is detecting and reporting attacks and invaders. One of the benefits of IDSs against other security methods is to cover a wide range of attacks on WSNs. Researchers have provided various IDSs for WSNs, but due to the limitations in WSNs, designing an effective and efficient IDS that be usable in them is still a major challenge. We plan to propose an efficient architecture for IDSs on WSNs by examining existing methods. Our meaning of efficient architecture is improvements in energy consumption and waste as the main parameter in sensor nodes as well as improving the detection accuracy of attacks. The main idea of our proposed architecture is to pay attention to the importance level of the nodes and their sensitivity in the network and, on this basis, we will use effective detection algorithms at different levels. Our goal in this architecture is to cover network laxyer attacks and routing, which are the most common attacks in sensor networks. In order to proper and desirable validation, by performing simulations of the proposed architecture, all performance criteria have been evaluated on that. The results obtained from simulations show that the proposed architecture is well-known as an effective and lightweight method for the WSNs, and by applying it in WSNs, the performance of the network can be kept in the optimum level.