Intelligent Selection of Requirements for a Secure Design of Smart City Transport Infrastructure Interfaces

The paper is devoted to the actual topic of ensuring the information security of the Smart City. Particular attention is paid to its subsystem – the transport infrastructure. An aspect of consideration is the security of interaction interfaces (implemented using separate hardware and software devices) used in the infrastructure. For this, the previous results obtained by the authors are used, namely, the structural model of the transport infrastructure. A semi-automatic scheme for the formation and ranking of security requirements for transport infrastructure interfaces at the stage of architecture design is described. To automate the determination of the parameters of new potential vulnerabilities, machine learning methods are used for danger level classification. The input data for the classifier is a textual human description of the vulnerability. Experiments are underway to classify vulnerabilities from the FSTEC of Russia database. A performance comparison is made for the following classifiers: ExtraTreeClassifier, Svm.SVC, MLPClassifier, ExtraTreesClassifier, RandomForestClassifier. The quality of classification expressed in the F-measure averaged \({\sim }\) 0.65, which can be considered satisfactory. However, subsequent experiments with the English-language NVD have already obtained a high value of the F-measure as \({\sim }\) 0.89. Obtained results showed the applicability of the proposed method for vulnerability assessment of Smart City transport infrastructure interfaces. The main contribution of the investigation is the vulnerabilities ranking method of according to its human-oriented description.