In recent years, the popularity of mobile terminal devices has promoted the wide application of biometric identification technology, but unprotected biometric templates face serious security and privacy risks. Moreover, mobile terminal devices have limited computational power and storage resources, which make it difficult to support very complex operations. To address the above problems, this paper proposes a lightweight cancelable biometric template protection scheme for mobile terminal devices. Firstly, the original biometric feature vector is copied and the copied feature vector is locally sampled using a random sampling sequence. Then, a random sampling-based grouped MinHash algorithm is proposed to transform the real-valued feature vector into binary hash codes. Finally, the lightweight encryption of biometric templates is achieved using a user-specific key and the XOR operation. Experimental results and analysis show that this scheme significantly reduces the computational overhead and storage overhead when generating biometric templates, and is highly universal for the protection of biometric templates, such as faces and fingerprints. Meanwhile, this scheme meets the design criteria of cancelable biometric template protection schemes, and is resistant to common security and privacy attacks.