International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056 Volume: 03 Issue: 09 | Sep-2016 www.irjet.net p-ISSN: 2395-0072 © 2016, IRJET | Impact Factor value: 4.45 | ISO 9001:2008 Certified Journal | Page 486 Privacy Preserving Classification over Semantically Secure Encrypted Relational Data in Cloud Environment Mr. Gaikwad Vijayendra Sanjay 1 , Dr. Khan Rahat Afreen 2 1 PG student, Deogiri Institute of Engineering and Management Studies, Dr. Babasaheb Ambedkar Marathwada University, Aurangabad, Maharashtra state, India. vij711@gmail.com 2 Associate Professor, CSE Department, Deogiri Institute of Engineering and Management Studies, Dr. Babasaheb Ambedkar Marathwada University, Aurangabad, Maharashtra state, India. rahatkhan@dietms.org ----------------------------------------------------------------------***--------------------------------------------------------------------- Abstract— The Cloud environment, with its extensive resources has become a good choice for organizations to keep their data and access it on demand. When the organizationǯs need is to just upload their data and use it as and when required from the cloud, the cloud service itself encrypts that data with its own credentials or in some cases, for maintaining the confidentiality, the data owners encrypt their data prior to outsourcing it. But there is no provision for processing some data within the cloud environment and at the same time maintain data confidentiality and privacy of user query. As a consequence, for classification, either the data needs to be decrypted by the cloud at some point of time and then processed to take proper classification decision or the data owner has no choice but to perform the same task at his/ her end partially or fully. Since the data used for classification is encrypted and placed onto a cloud, the conventional privacy preserving classification methods are not suitable. Some recent work has been done in this direction, but it is proven to be computationally costly and also not very practical. Our proposed system is an effort towards resolving this very problem of classifying encrypted user queries over encrypted data in a more effective and time efficient manner. This is achieved re- designing the existing privacy preserving protocol from a different perspective and by leveraging the properties of homomorphic cryptosystem. Our approach is computationally inexpensive and does not compromise the privacy of user query or the confidentiality of the database outsourced by the data owner. Keywords— encrypted database, homomorphic cryptosystem, k- nearest neighbors, security 1. INTRODUCTION The recent trends in cloud services have revolutionized the outlook of organizations towards leveraging the benefits of outsourcing their data. Cloud computing, with its platform as a service (PaaS) feature, has seriously grabbed the attention of organizations desiring to completely outsource their valuable data along with the some data management tasks. But, despite of various facilities that cloud avails, there are still some data confidentiality and privacy issues that keep the organizations from utilizing them. When data is straight away uploaded to the cloud, the cloud itself encrypts it, for securing it from any third party theft and then stores it. By doing so, the data is open for the cloud service providers at the first place which can be threat. If the data contains very sensitive information such as medical records of patients, then somewhere down the line the patientsǯ privacy gets compromised. To avoid this, the first solution that organizations use is to encrypt their data, prior to uploading it to the cloud. But what when the use of this data is just not limited to its retrieval? To perform some processing over this encrypted data at the cloud without ever decrypting it is very difficult task. The privacy issues involved in this kind of situations can be explained by the example. Consider that a hospital keeps their patients encrypted database on cloud along with the data mining task. Now, when a doctor wants assert about symptoms of a disease of the patient, which he/ she cannot affirmatively treat, the doctor can use relative classification process and find out the disease with which the patient is suffering. For getting a precise response, the doctor needs to trigger a query for the classification process on cloud, which would contain patientǯs highly personal information. So, it is very obvious that this query must be encrypted prior to sending it to the cloud, in order to protect the patientǯs privacy. Thus, it is important to consider the privacy of the usersǯ query when it is involved in the data mining task. Also, any cloud malfunction activity can determine useful information about data access patterns although data are always encrypted. Therefore, we can say that, while performing a classification or any other data mining task on encrypted data in an outsourced environment as cloud, the data ownerǯs confidentiality, user queryǯs privacy and preventing the cloud from learning any access patterns must be the foremost objectives. In this paper, we have proposed some methods which collaboratively solve the secure classification over encrypted data problem assuming that encrypted data and the classification process are outsourced. Although each of