This paper addresses the biometric security problem on smartcard and proposes a new method to protect face biometric data against attack. The proposed method is developed based on the Reed-Solomon codes, with two modifications. First, to handle the image variations problem, a bounded distance encoding algorithm is used to reduce the within-class distance. Second, since the dimension of the face feature vector is relatively large, we divide the feature vector into smaller segment. In this way, we found that it will not only affect the performance, but also the security of the encoding algorithm. A theoretical analysis is also given in this paper. Two most popular appearance-based methods, namely, Eigenface and Fisherface, are selected to construct the feature vector and ORL database is used for evaluation. In Eigenface system, the proposed algorithm offers a security level of 120 bits, with an acceptable error rate of around 3%. In Fisherface system, the proposed algorithm offers a security level of 62 bits, with equal error rate around 5%. The results are comparable with performance of Eigenface and Fisherface without protection.