Traditional filtering rules based on network and packet attributes have proven insufficient in dealing with the evolving landscape of network security threats. This inadequacy not only jeopardizes the sustainable security governance of network data but also underscores the need to characterize the content of network data for a deeper understanding of its security posture. With the ongoing advancement of technologies like cloud computing and big data, the extensive content feature-matching required for network data filtering has become a performance bottleneck in network security governance. As a solution to this challenge, this paper proposes a feature-matching algorithm for web data based on word vector space indexing (FWVS). The algorithm employs a Skip-gram model to transform words within the training corpus into word vectors. Subsequently, it constructs an n-dimensional spherical space for the target words based on these word vectors. An iterative algorithm is introduced to establish the minimum Euclidean space encompassing all the target words. Finally, the text to be retrieved is segmented into words, and a comparison is made between the n-dimensional spherical space and the retrieved words to determine their inclusion. The experimental results demonstrate that for a text size of 800KB, the FWVS algorithm's retrieval speed is 1.2 times that of the ACBM algorithm, 1.7 times that of the WM algorithm, and 2.8 times that of the AC algorithm.