Cancelable biometrics methods can be unimodal or multimodal. Unimodal cancelable biometric methods use single biometric traits, whereas multimodal methods use more than one. Unimodal methods are less expensive and require less cost and storage than multimodal methods, but multimodal methods are more accurate and secure. This paper applies existing random permutation and dimensionality reduction-based methods template generation methods for generating multimodal biometrics cancelable templates. The fusion of Georgia Tech face and CASIA-IrisV1 biometrics creates the multimodal biometric database. The performance analysis shows that these unimodal methods can be applied to multimodal biometric applications and improve/retain recognition accuracy. The templates generated using the fusions are also irreversible and immune to different impostors’ attacks as evaluated using the Unified Average Changing Intensity (UACI) and Number of Pixel Change Rate (NPCR). If the templates are compromised, they can be cancelled and replaced with new ones that do not correlate with the old ones.