Robust insider attacks countermeasure for hadoop: Design and implementation

Hadoop is an open source software framework for storage and processing of large-scale datasets. The proliferation of cloud services and its corresponding increasing number of users lead to a larger attack surface, especially for internal threats. Therefore, in corporate data centers, it is essential to ensure the security, authenticity, and integrity of all the entities of Hadoop. The current secure implementations of Hadoop mainly utilize Kerberos, which is known to suffer from many security and performance issues, including the concentration of authentication credentials, single point of failure, and online availability. Most importantly, these Kerberos-based implementations do not guard against insider threats. In this paper, we propose an authentication framework for Hadoop that utilizes trusted platform module technology. The proposed approach provides significant security guarantees against insider threats, which manipulate the execution environment without the consent of legitimate clients. We have conducted extensive experiments to validate the performance and the security properties of our approach. The results demonstrate that the proposed approach alleviates many of the shortcomings of Kerberos-based state-of-the-art protocols and provides unique security guarantees with acceptable overhead. Moreover, we have formally proved the correctness and the security guarantees of our protocol via Burrows-Abadi-Needham logic.