This document analyzes the effect of compressive sensing theory and watermarking on verification and authentication performance in a multibiometric system. It proposes embedding sparse measurements of a watermark biometric image, generated using compressive sensing theory, into transform coefficients of a host biometric image. This provides two levels of security and authentication using the watermarked host image and reconstructed watermark image. Experimental results on different watermarking techniques show the proposed techniques do not affect the multibiometric system's verification performance or authentication accuracy while providing security against various attacks.