The network plays a crucial role in a wide range of applications in the era of digital transformation. However, this development is limited by multiple challenges, and security problems are the most dangerous ones due to their huge negative impacts. The Intrusion Detection System (IDS) is a method that can detect network activities early, but the construction of an efficient IDS is complicated. Therefore, RWIDS, an approach feature selection method focusing on the network security domain, is proposed, which associates the strength of the graph for the presentation of network data and the Random Walk algorithm to find the optimal feature set. The effectiveness of RWIDS is proven through the benchmark with the 5-fold cross-validation procedures and the comparison stage, which indicates that with only Top 15 the most important features, the performance and the processing time are improved considerably. As a result, RWIDS shows the potential of reducing the dimensions of network data, which possibly contributes to building an efficient IDS for minimizing huge losses more effectively.