The context of multibiometric plays a pivotal role in enhancing an identification system, since a biometric system is now the most physical way of identifying and verifying individuals. The feature of multibiometric could be merged to produce identification information. However, unimodal biometric systems suffer from different types of breaching. Thus, mixing biometrics with cryptography leads to overcome small variations existing between diverse acquisitions of the same biometric in order to produce the robust system. In this paper, a new robust multibiometric system is proposed to create a random key from person multibiometric, facial and fingerprint images which are used simultaneously to produce this key. Several manipulations are made on compactness information for these two images to get a unique key for each person. The generated random key can be used for electronic numbers, passport identification, civil identification card, and it could be used as seeds for pseudo-random number generators. The multi-biometric system operates on two images, faces and fingerprints, by partitioning each image into four parts and taking the highest density for each one, XOR these parts; diffusions process is applied on these parts including permutation and thresholding to produce a random key. The generated key cannot be revocable that passed through randomness tests to ensure whether the generated key is accepted as true. Thus, the results of the tests are passed and presented that all generated keys are accepted to be random and unpredictable binary sequences and hence they can be used efficiently.
