Research On Template Protection Methods Applied To Face Identity Features

Biometric identification technology has been widely applied in the field of identity authentication,especially face recognition technology,which has become one of the most mature biometric identification technologies with its advantages of low cost,friendliness and high acceptance.However,face recognition technology needs to save the user’s face feature template into the database,and in real applications,the security of this storage method cannot meet the needs of complex network environments,so the protection of face feature templates has become one of the important issues in face recognition technology.This thesis focuses on the protection of face feature templates in face recognition systems,and investigates the template protection method applied to face identity features based on the construction of a face recognition system consisting of face image pre-processing,feature extraction and template matching.For the problem of small key spaces and dependence on initial values for Logistic Chaos Mappings,an improved Logistic Chaos Mapping face template protection method incorporating face feature information is studied.The method iterates the improved Logistic chaos formula with the feature vector for a number of times of equal length to generate a chaotic sequence,adjusts the chaotic sequence using the extracted feature vector,generates a new iteration of the initial value to obtain a new sequence,and dissociates the sequence with the feature vector to obtain an encrypted template.The experimental results show that the key space of the method reaches 10⁴⁵×255,which has better template protection performance compared with the original Logistic chaotic mapping.The recognition rate of the method improved by 0.65%and 1.5%compared to the original Logistic Chaos Mapping in both the ORL and Yale datasets respectively.To further improve encryption performance and expand the key space.Study of an improved Logistic Chaos Mapping face template protection method incorporating face feature information.On top of the improved Logistic Chaos Mapping that fuses the face feature information to encrypt and change the feature values,the resulting encrypted templates are then position scrambled.The first encrypted template is chunked to obtain two order matrices.After the Arnold permutation of the two matrices,the two templates are stitched together to obtain a face feature template with two encryptions.The experimental results show that the method improves the encryption performance based on the improved Logistic Chaos Mapping and the key space is 10⁵³×3.The correct recognition rate of the method was 94.65%and 93.22%in the ORL and Yale datasets respectively,which can meet the requirements of face recognition.The research in this thesis achieves the protection of face identity templates,meeting the requirements of irreversibility,revocability and unlinkability for biometric template protection.The security performance of the protection of face identity templates is satisfied while ensuring the correct recognition rate of face recognition.