Design and Implementation of Data Encryption in Cloud based on HDFS

Abstract

Abstract-As an open-source distributed programming framework, Hadoop has gradually become popular in industry recently. Its distributed file system (HDFS) enables storing large data with advantages of high fault tolerance and throughput. However, the fact that the current HDFS does not support intra-cloud data encryption yet makes data privacy becomes a key security issue. This paper presents ahybrid encryption method based on HDFS. We adopt symmetric encryption to encrypt and decrypt file blocks at datanodes and use asymmetric encryption scheme to protect the symmetric keys. By this method, we can prevent datanode intruders from stealing user data, while ensuring that clients are lightweight. The experiments show that with and without block encryption algorithm, our solution brings43% and 2% performance degradation compared to the generic HDFS.

An Integrated Methodology for Big Data Classification and Security for Improving Cloud Systems Data Mobility

“…The experiments indicated that HDFS performed best when big data is partitioned into smaller file sizes. It can be inferred from Figures 3, 4, and 5 that our securing algorithm outperforms its counterparts in [32] and [46] for response time, delay time, and throughput.…”

“…User authentication at the sender and the receiver clouds are carried out by integrating a user private key with a public key. Our security experimental results are compared with non-secured Hadoop baseline [4] and the secured methods proposed in [46] and [32].…”

“…The method proposed in [46] presents a hybrid encryption method to protect file blocks and session keys based on HDFS. Symmetric encryption technique is adopted to encrypt and decrypt file blocks at clouds data nodes while asymmetric encryption technique is used to secure the symmetric keys.…”

“…This key serves as an authentication to both clouds that enable them to generate and verify the necessary tokens without holding them for transferring data from one cloud to another. Therefore, our approach decreases the unnecessary network bandwidth and processing overhead caused by other approaches in comparison [32] and [46] due to extra transmission, encryption, and decryption operations between the sender and the receiver clouds. Hence, significance decreases in the total transmission response time, the total delay time, and the throughput degradation are achieved by our protocol which results in better HDFS security performance.…”

“…The big data security approach described in [46] proposes a data transmission technique that depend on secure socket layer (SSL) connection between the name nodes of both sender and receiver cloud systems. The cloud receiver name node generates a temporary session key, a random hash number and series of tickets encrypted by the session key to communicate with the cloud sender data nodes, compute the double hash value, and return the encrypted tickets to the sender name node respectively.…”