We propose a composite signature-based digital watermarking technique for fingerprint verification and identity extraction applications. In a two-step process, our scheme attempts to integrate three modalities of authentication-"what you know," "what you have," and "what you are." Using a signature extracted from the Fourier phase of an original image, we hide an encoded signature back into the original image using a frequency domain bit-plane embedding technique. Additionally, other identity information is embedded in the spatial domain using a spread-spectrum method. The detection process computes the Fourier transform of the watermarked image, then extracts the embedded signature and correlates it with a calculated signature. The embedding of a composite signature ensures robustness against any chosen form of distortion. Simulation results for the signature and identity embedding and detection are provided for the Fingerprint Verification Competition (FVC) fingerprint image database.