Data Processing Under Privacy Protection

With the development of the Internet,privacy issues have received more and more attention,but in the field of privacy protection data processing,there are still problems to be solved.The cloud server has a wealth of computing resources and can perform the specified computing tasks.With the rapid growth of personal and corporate data,more and more computing tasks are difficult to handle locally,and they are outsourced to cloud servers for computing.If our data is directly exposed to the cloud server,we will directly face the problem of privacy leakage,how to protect the private image data while implementing image computing has become our main focus.In order to solve this problem,this paper studies the processing and feature extraction of privacy protection data based on homomorphic encryption scheme encryption.Based on the previous research results,this paper studies the use of data encrypted based on homomorphic encryption scheme in neural network and the feature extraction of privacy image based on homomorphic encryption scheme.The research results of this paper mainly include the following contents:Using homomorphic encryption to construct a neural network that can process privacy protection data,an excitation function that can support the properties of homomorphic operations is designed.Analysis of the algorithm shows that our solution can complete the calculation in polynomial time.A scheme for extracting secure Haar features on privacy protected images is proposed.Based on the homomorphic encryption scheme,secure Haar feature extraction on the image ciphertext domain is implemented.The secure Haar feature is the performance of the Haar feature on the ciphertext domain.Compared with the original Haar feature,it can effectively protect the privacy of the image while maintaining validity.Through the analysis and evaluation of the algorithm and the eigenvalues,the secure Haar feature obtained by our scheme is very similar to the Haar feature extracted from the plaintext image and can be completed in polynomial time.A scheme for numerical comparison of dense states is proposed.We use the homomorphic nature of homomorphic encryption to design a dual encryption scheme,which involves two parties,and can obtain comparison results with very low computing resources in the case of one interaction.Apply secure Haar features.The face detection of the privacy-protected image is realized by using the secure Haar feature extracted in the image ciphertext domain.By analyzing the experimental results,the detection result of our scheme is very similar to the original Haar feature detection result.