High-dimentional data authentication protocol based on hash chain for Hadoop systems

Abstract

Social media networks along with mobile devices are major sources of big data. A recent trend involves using Hadoop for big data to extract unique trends, insights, and patterns. However, Hadoop lacks secure authentication mechanisms that prevent unauthorized access to data. The current authentication token system that is used for user authentication in the Hadoop distributed file system (HDFS) is vulnerable to replay and datanode hacking attacks. This paper proposes a data authentication protocol that protects sensitive data in HDFS against malicious attacks. The proposed protocol uses the hash chain method instead of the key exchange method based on public-key encryption. An HDFS blocks in datanodes of the proposed protocol are accessed by a client with a valid block access token from a namenode. In addition, clients' anonymity is guaranteed because an HDFS adopting the proposed protocol passes hash-chained block values so the HDFS blocks that generates security awareness information is not exposed unnecessarily to a third party. An HDFS adopting the proposed protocol has the performance (e.g., communication power, computational power, and area efficiency) as good as that of conventional HDFS systems.