The key to the difficulty of assessing the value of leakage data is that the methods of obtaining the value characteristics of streaming data through full analysis are very time-consuming and the value assessment dimensions of leakage data are difficult to describe. Therefore, based on the idea of mineral drilling exploration, we propose a SDSLA method, which can retain more valuable discrete values under the constraint of limited access, so that the sample can accurately assess the value characteristics of the streaming data. We further propose a MIRS algorithm to make the sample and streaming data have the same class basic imbalance rate, and a value assessment model to comprehensively assess the value of leakage data from 6 aspects. The experimental results demonstrate that the SDSLA and MIRS can well assess the value characteristics of the streaming data, and each index is almost above 90%, which is superior to the classical reservoir sampling. It is observed that our proposed methods and model can help to efficiently assess the value of leakage data under non-full analysis.