A Sentence-BERT-based Model for Expressing Key Features of Hospital Web Logs

Hospital web application log data contains a significant number of specialized terms, and there is a high degree of similarity in their expressions and content. This similarity often leads to a high false alarm rate in hospital network security detection. In this paper, we propose a SB-KFR model (Sentence-BERT-based Key Feature Representation) to tackle this problem. This model converts hospital web logs into feature vectors by extracting key features and performing vector transformation. In this paper, seven machine learning models are used to verify the feature vector. The experimental results demonstrate a reduction in false positives for hospital web application intrusion detection after applying the SB-KFR model to process the web logs.