Research on the construction and optimization of information security risk assessment model based on intelligent algorithm

Traditional risk assessment methods rely on expert experience and manual analysis, and it is difficult to deal with complex and changeable security threats. In this paper, an information security risk assessment model construction and optimization method based on intelligent algorithm is proposed. First of all, comprehensively identify the risk factors faced by the information system, including technical loopholes, network attacks, user behavior, etc., and build an evaluation index system covering threat identification ability, vulnerability assessment, impact analysis and risk calculation. Next, a Deep Neural Network (DNN) structure is designed, incorporating an attention mechanism to enhance the learning capability of key features. A hybrid ensemble optimization strategy is employed to construct a Stacking model architecture, integrating base models such as DNN, Support Vector Machine (SVM), and Random Forest (RF). Model fusion is achieved through Gradient Boosting Decision Trees (GBDT). Meanwhile, adaptive regularization methods and a multi-objective optimization function are introduced to further enhance the model's performance. Taking a provincial data center of a bank as an example, the data from 2020 to 2024 are used for model training and testing. The results show that the AUC-ROC value of the mixed model is 0.95, the F1-Score is 0.88, and the recall rate of high-risk events is 93.7%, which is significantly better than the single DNN model and SVM model. The dynamic regularization strategy effectively improves the over-fitting problem of the model, and the application of multi-objective loss function further improves the model performance.