Attack method for big data processing platform Hadoop

Abstract

The invention belongs to the field of big data security in the field of computers, and relates to an attack method for a big data processing platform Hadoop. The attack method is provided with a speculative task scheduling attack model, a heartbeat attack model, a data attack model and a Shuffle attack model, wherein the speculative task scheduling attack model can slow down the execution of individual tasks for speculative task scheduling algorithm attacks in a cluster, and then slow down the execution of the whole operation; the heartbeat attack model prolongs heartbeat sending time to timeout time for attacks of a heartbeat mechanism in the cluster so as to influence performance of the cluster; the data attack model attacks intermediate data in the mapreduce task execution process, andcarries out operations such as modification, deletion and addition on the content of an intermediate result so as to influence the whole job execution result; and the Shuffle attack model attacks thecluster at a Shuffle stage in the task execution process, and prolongs the execution time of the whole task by prolonging the execution time of the Shuffle stage.

Description

technical field [0001] The invention belongs to the field of big data security in the computer field, and relates to an attack method for Hadoop, a big data processing platform. Background technique [0002] With the rise and development of cloud computing and big data, more and more enterprises and users use big data processing platforms such as Hadoop and Spark to process and analyze massive data. For example, platform-as-a-service (PAAS) platforms provide application development and runtime environments through cloud computing technology. Since the results of big data analysis provide critical information for a large number of applications, the security of data processing frameworks becomes a very important issue. [0003] Existing big data processing platforms are distributed platforms. All operations are not executed on one machine, but are distributed on different machines, which can shorten the execution time of jobs and improve execution efficiency. However, due t...

AI Technical Summary

Problems solved by technology

[0005] The paper "Fu X , Gao Y , Luo B , et al. Security Threats to Hadoop: DataLeakage Attacks and Investigation[J]. IEEE Network, 2017, 31(2):67-71." Although it has detailed data leakage attacks analysis, but did not give a specific implementation of the attack, and did not compare it with other types of attack schemes

Paper "Wang J , Wang T , Yang Z , et al. SEINA: AStealthy and Effective Internal Attack in Hadoop Systems[C] / / International Conference on Computing, Networking and Communications (ICNC 2017). IEEE,2017: 525-530." Proposed an attack scheme inside Hadoop, but it also did not analyze other security vulnerabilities, and compared it with other attack schemes

Images