Sensor networks are composed of a large number of low power sensor devices. For secure communication among sensors, secret keys must be established between them. Recently, several pairwise key schemes have been proposed for large distributed sensor networks. These schemes randomly select a set of keys from a key pool and install the keys in the memory of each sensor. After deployment, the sensors can set up keys by using the preinstalled keys. Due to lack of tamper-resistant hardware, the sensor networks are vulnerable to node capture attacks. The information gained from captured nodes can be used to compromise communication among uncompromised sensors. Duetal. , Liu and Ning  proposed to use the known deployment information to reduce the memory requirements and mitigate the consequences of node capture attack. Our analysis shows that the assumption of random capture of sensors is too weak. An intelligent attacker can selectively capture sensors to get more information with less efforts. In addition to selective node capture attack, all recent proposals are vulnerable to node fabrication attack, in which an attacker can fabricate new sensors by manipulating the compromised secret keys and then deploy the fabricated sensors into the sensor system. To counter these attacks, we propose a grid-group scheme which uses known deployment information. Unlike the pairwise key scheme using deployment information proposed by Du et al, we uniformly deploy sensors in a large area; instead of randomly distributing keys from a large key pool to each sensor, we systematically distribute secret keys to each sensor from a structured key pool. Our performance analysis shows that our scheme requires less number of keys preinstalled for each sensor and is resilient to selective node capture attack and node fabrication attack.