.png){kind=icon}
.png){kind=icon}
Return
Verifiable speech retrieval algorithm based on KNN secure hashing
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-022-13387-w.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
With the rapid development of mobile Internet, the dimension of speech data is too high and the space is complex. The existing speech retrieval algorithms can not meet the efficient retrieval efficiency and privacy security of speech data in massive applications. Aiming at the problems of retrieval efficiency and accuracy caused by high dimension and complex space of speech feature data, content verifiable retrieval after speech attack, and the security of speech storage and transmission process, a security framework based on KNN Secure Hash (KNNSH) is proposed for verifiable speech retrieval. In this algorithm, the spectral centroid of speech is used as the only input factor, and then KNN classification is used to train and learn the speech vector to obtain each speech centroid. Each speech centroid is assigned a specific hyperchaotic Lorenz compressed sensing encryption algorithm (HL-CS) key, and the security framework is constructed according to the revocable biometric template generated by the combination of classification and specific key. The binary hash vector is generated, and then the hash vector is encrypted by HL-CS. The same encryption algorithm is used to encrypt the original speech. Experimental results show that only one item needs to be matched in the intra class matching process after classification, which improves the retrieval efficiency and accuracy, and realizes the content verification of speech retrieval after content preservation operations. Speech encryption effectively prevents the disclosure of plaintext, ensures the security of speech storage and transmission process. It has a large key space, which is enough to resist exhaustive attacks.
Keywords:
Verifiable speech retrieval
KNNSH
HL-CS
Biometric template
Security
Journal
IF:
3
Papers:
1.9W
Citations:
3.2W