The authentication based on biometric is more reliable and secure because of using the unique physical feature of human. Initially, mono biometric system was used for authentication but it has some error rate and hence multimodal biometric system was introduced to reduce the error. The constraint of using these systems is to maintain more information. Without reducing the error rate and maintain the above constraint a new algorithm has been developed which is based on the invariant moment information of fingerprint and face which is fused using variation. In this algorithm the fingerprint and face is segmented and the invariant moment information is extracted. The invariants are fused into a single identification value by using coefficient variance. This single value is authenticated by calculating the difference, evaluated using the threshold value which is set as 90% for fingerprint and 70% for face, provides low error rate of FAR and FRR. The algorithm is tested under cooperative and non cooperative condition and obtained less complexity, storage, execution time, high reliability and secure authentication system. © 2010 Springer-Verlag Berlin Heidelberg.