Biometric recognition is a broad and dynamic field of research but the main problem of this field is spoofing attack or presentation attack "use of fake biometric in place of the real biometric sample from original user". Liveness detection is the prime countermeasure to spoofing attacks, which is based on the principle that some additional information can be obtained to verify that the produced data is genuine or not by a standard verification system. It utilized anatomical signs of life, such as facial expression, blinking of eyes, movement of the head, etc. This paper presents a comprehensive review of various liveness detection techniques based on a multimodal biometric system, in which physiological and behavioral properties are used to differentiate between genuine and fake biometric traits. Multimodal systems utilize two or more biometric traits which makes them more secure as compared to unimodal systems. These systems overcome the limitations of the unimodal system such as spoof attack, noisy data, non-universality, distinctiveness, and intra-class variations, etc. Hence to make the biometric systems more secure and robust, multimodal techniques are used. In this paper, we categorized and discuss the various multimodal biometric techniques proposed by various researchers in the last decade, and a new classification is also developed for the same. This paper covers theories, methodology, evaluation datasets, and aims at future work in this field of research.