We present a novel information theoretic approach to make network coding based storage secure against pollution attacks in sensor networks. The approach is based on a new decoding algorithm which makes it possible to find adversarial blocks using one more encoded block than strictly necessary for decoding. Our scheme fits well to the requirements of sensor networks, because it operates with adding very low computational and communication overhead to source and storage nodes, only the collector node needs to perform some additional computation. Our approach does not apply cryptography, hence it works in environments where no pre-shared keys, secure channels or PKI are available, which is often the case in sensor networks.