Pervasive computing is changing the life style of human being in the 21st century. Smart home is one of the emerging examples for pervasive computing applications. In a smart home scenario, Wireless Sensor Network (WSN) can potentially provide information about device activities and their status. This information can be useful for variety of purposes like monitoring home security, examining status of nodes, and replacing nodes that are malfunctioning or dead. The sensor network needs to be protected from intrusions and anomalies. Over the years, various Intrusion Detection Systems (IDS) are proposed for preventing WSNs from intrusions. Although some mobile agent based IDS are proposed, anomaly detection for wireless smart home sensor network (WSHSN) is not fully exploited. In this article, we have proposed a mobile agent based anomaly detection mechanism for WSHSNs. The proposed scheme takes advantage from heterogeneous nature of devices in smart home for effectively detecting anomalies. The anomaly detection infrastructure is installed at resource rich nodes such as Cluster Heads (CHs).The mobile agent takes only few bytes of data to further assess the anomalous behavior of suspicious node. The analysis and comparative study show several advantages of the proposed scheme, such as efficient utilization of memory, reducing network load, and reduction in overall computational cost of the network. The nature of our proposed solution is generic, so that as it is, or with slight modifications, it can be implemented on other heterogeneous wireless sensor networks.