The invention discloses a service denial detection method for a large data platform. The method includes the steps that a plurality of nodes are selected in the large data platform and serve as detection nodes, and a plurality of nodes serve as calculation nodes; a user blacklist is set; secondly, a submitted application is firstly detected in the detection nodes; if any task is detected unqualified, the task and the application are both marked failed; if tasks are detected qualified, the application is allocated to the calculation nodes for continuous execution; thirdly, the nodes are detected, and when the number of the tasks marked failed on the nodes exceeds a set value P, the nodes are marked UNHEALTHY; when the number of the calculation nodes marked UNHEALTHY exceeds a set threshold, the tasks marked failed on the calculation nodes are detected, and if the tasks come from a same user or the similarity exceeds a set value, the calculation nodes are added into an exclusion list; if the ratio of the number of the nodes in the exclusion list exceeds N%, it is judged that attack of service denial exists.