期刊名称：International Journal of Innovative Research in Science, Engineering and Technology
印刷版ISSN：2347-6710
电子版ISSN：2319-8753
出版年度：2015
期号：MULTICON
页码：336
出版社：S&S Publications
摘要：Data categorization methods are used to assign class labels to the transactional data values. Resourcerequirement for the data categorization process is very high. In cloud environment users’ data are usually processedremotely in unknown machines that users do not own or operate. User data control is reduced on data sharing underremote machines. Anomalous and normal transactions are identified using classification techniques. Neural networktechniques are used for the classification process. Back-Propagation Neural network (BPN) is an effective method forlearning neural networks. Input layer, hidden layer and output layer are used in the neural network operations.Shared data values are maintained under different parties to perform the data categorization process. A trusted authority(TA), the participating parties (data owner) and the cloud servers entities are involved in the privacy preserved miningprocess. TA is only responsible for generating and issuing encryption/decryption keys for all the other parties.Participating party is the data owner uploads the encrypted data for the learning process. Cloud server is used tocompute the learning process under cloud resource environment. Each participant first encrypts their private data withthe system public key and then uploads the ciphertexts to the cloud. Cloud servers execute most of the operations in thelearning process over the ciphertexts. Cloud server returns the encrypted results to the participants. The participantsjointly decrypt the results with which they update their respective weights for the BPN network. Boneh, Goh andNissim (BGN) doubly homomorphic encryption algorithm is used to secure the private data values. Data splittingmechanism is used to protect the intermediate data during the learning process. Random sharing algorithm is applied torandomly split the data without decrypting the actual value. Secure scalar product and addition operations are used inthe encryption and decryption process.The privacy preserved data categorization scheme is composed without the trusted authority for key managementprocess. Key generation and issue operations are carried out in a distributed manner. Cloud server is enhanced to verifythe user and data level details. Privacy preserved BPN learning process is tuned with cloud resource allocation process.
