The multi-tenant cloud computing infrastructure has become popular because many service providers can deliver their services through it. In dataflow processing applications, the dataflow need to be sequentially processed by a lot of processing components that are provided by different service providers. In order to prevent malicious attackers from modifying processing results or providing incorrect functions, dataflow processing integrity must be ensured. In this paper, we propose RandTest, which is a lightweight dataflow integrity checking method. RandTest is a probabilistic integrity attestation method, which is very cost-efficient. By properly choosing parameters, the dataflow integrity can be guaranteed with a large probability. We have both theoretically analyzed and experimentally measured the performance of the RandTest method. The results show that RandTest is very suitable for multi-tenant cloud computing.