Abstract: In the recent cloud era, computing moves to a new plane of running large scale scientific applications. Many parallel algorithms have been created to support a large dataset. MapReduce is one such parallel data processing framework adopted widely for scientific research, machine learning and high end computing. The most prevalent implementation of MapReduce is the open source project Hadoop. To protect the integrity and confidentiality of data uploaded, MapReduce introduced a Kerberos-based model with tokens for datablocks and processing nodes. The tokens are symmetrically encrypted and distributed across the nodes. Such a technique is vulnerable to man-in-the-middle attacks like data loss, data modification and stealing of keys. In this study, a novel technique is proposed based on public key encryption on top of the Kerberos model to enhance Security. The various attack scenarios on the current Hadoop implementation model has been analyzed and a secure environment has been proposed. The study shows that the proposed framework provides an improved level of security when using RSA (Rivest Shamir Adleman) with 65,537 keysize consumed 23 milli seconds, while using 257 bits keysize which consumed 21 milli seconds.