This paper proposes a Encrypted speech retrieval based on long sequence Biohashing to solve the problem of plaintext data leakage in the existing speech retrieval system, and improve the efficiency and accuracy of speech retrieval, the diversity and revocability of biometric template. According to speech feature classification, a biometric template with a single mapping key is established, and then the feature vector is used to generate speech feature index, and the speech file is encrypted by the improved SHA256 algorithm, and finally, feature index and encrypted speech are sent to the cloud. For the input speech, the feature vector is generated at the mobile, and then the cloud will retrieve the feature vector table according to the feature vector to obtain the feature index of the speech, and finally, the feature index only matches the feature index related to the speech in the feature index table. Experimental results show that this algorithm can not only effectively prevent the leakage of plaintext, but also has good diversity and revocability of biometric template. At the same time, the algorithm not only has good efficiency and accuracy, but also solves the problem of speech retrieval after content preservation operation.