| Summary: | 碩士 === 國立高雄第一科技大學 === 電腦與通訊工程所 === 95 === The techniques for identity recognition by using human face have become more popular in the applications of e-commerce. When a customer shops online, buyer’s face image taken by a camera in real time. However, when the face image is transmitted on the network, it might be intercepted by a hacker, and modified or reserved for later use in fake transactions. In this thesis, three distinct methods for security protection of online transmitted face images is proposed, which may be selected for different applications.
The first method adopts information hiding techniques to embed information of the e-commerce transaction, compressed facial region, and authentication signals into the face image excluding its facial region. When the server receives a transmitted face image, it can then extract not only the information of the transaction but also judge the truthfulness of the image. If facial region of face image was modified, the original facial region can be reconstructed. In this thesis, we adopt an adaptive JPEG quantization table which decided by the embedded data size and the DCT coefficients of the face image in order to decrease the influence of image quality caused by the embedding steps. The second method compresses the DCT coefficients of the JPEG face image. After compressing, we can obtain space for embedding information of the e-commerce transaction, and authentication signals. This method produces a noise like image which is then transmitted to the server. Because the face is not exposed during the transmission process, this method can achieve additional degree of security. At the receiver end, the original face can be reconstructed by a reversible process. The third method uses a bit-shift operator for DCT coefficients of the JPEG face image. After bit-shift, we can obtain space for embedding information of the e-commerce transaction, and authentication signals. And we modify quantization table to keep image quality. At the receiver end, the original face can be reconstructed by a reversible process. Experiment results show the proposed methods are feasible.
|
|---|
