Phishing attacks have evolved in terms of sophistication and have increased in sheer number in recent years. This has led to corresponding developments in the methods used to evade the detection of phishing attacks, which pose daunting challenges to the privacy and security of the users of smart systems. This study uses LightGBM and features of the domain name to propose a machine-learning-based method to identify phishing websites and maintain the security of smart systems. Domain name features, often known as symmetry, are the property wherein multiple domain-name-generation algorithms remain constant. The proposed model of detection is first used to extract features of the domain name of the given website, including character-level features and information on the domain name. The features are filtered to improve the model’s accuracy and are subsequently used for classification. The results of experimental comparisons showed that the proposed model of detection, which integrates two types of features for training, significantly outperforms the model that uses a single type of feature. The proposed method also has a higher detection accuracy than other methods and is suitable for the real-time detection of many phishing websites.

Based on the NTRU trapdoor used in NIST’s Falcon, a signcryption scheme following the sign-then-encrypt paradigm is constructed. The existing partitioning technique based on Waters hash over the lattice can not complete the security reduction in the standard model for the signature part due to the “partiality” of the pre-image generated with the NTRU trapdoor. To address this, a variant of Waters hash over small integers is proposed and, the probability of the successful reduction is analyzed. The resulting signcryption achieves existential unforgeability under the adaptive chosen-message attacks. By utilizing the uniqueness of the secret and the noise in an NTRU instance, the tag used in encryption is eliminated. Furthermore, a method to construct tamper-sensitive lattice public key encryption is proposed. This approach implants the ciphertext-sensitive information into the lattice public key encryption and binds it to the encrypted information. The malleability to the public key ciphertext triggers the change of the message–signature pair so that the IND-CCA2 security of the entire ciphertext can be guaranteed by the signature for the message. Thanks to the rational design and the efficiency of the NTRU trapdoor, the computational overhead of the proposed scheme is reduced significantly compared to the existing lattice-based signcryption scheme, reaching orders of magnitude improvement in efficiency. The experiment shows that the proposed scheme is efficient.

This article proposes a prediction model that combines network physical characteristics and deep learning methods to solve the problem of insufficient round-trip time (RTT) prediction accuracy caused by complex dynamic characteristics in civil aviation business data networks. This model is based on a physical information neural network framework, which embeds domain knowledge such as delay load relationships, burst traffic attenuation patterns, and inverse RTT window constraints into a long short-term memory (LSTM) network. In addition, the multi constraint loss function enhances the adaptability of the model to complex network activities. Moreover, in order to overcome the limitations of LSTM networks in modeling long-range dependencies, a multi head attention (MA) mechanism was implemented to capture long-term temporal dependencies in parallel, thereby improving the model’s ability to capture long-range time step correlations. Benchmarking and extension tests were conducted on four civil aviation business data network datasets. The experimental results show that compared with the baseline model, the proposed model significantly improves the root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) for bidirectional network communication RTT prediction tasks. The study has verified that combining network physical attribute constraints with attention mechanisms can effectively improve the accuracy of transmission delay prediction, providing an effective method for effective traffic prediction in highly dynamic civil aviation network environments.

To address the challenges of heterogeneous desensitization requirements, limited system resources, and inefficient resource allocation for structured sensitive data, this paper proposes a data security-oriented multi-objective optimization method for desensitization resource allocation of civil aviation passenger sensitive data. The proposed method integrates heterogeneous resource scheduling with multi-objective optimization to achieve adaptive resource allocation under resource-constrained environments. A clustering-based approach is employed to group desensitization tasks with similar computational characteristics, and a personalized sensitivity coefficient is introduced to characterize task complexity and guide adaptive resource allocation. Based on task-group characteristics, a multi-objective optimization framework integrating the Non-dominated Sorting Genetic Algorithm II (NSGA-II) is developed to jointly optimize desensitization execution time, computational resource consumption, and task-group matching cost. Experimental results show that the proposed method successfully optimizes 92.76% of desensitization tasks, while 31.59% achieve optimal solutions with fewer thread resources than the original configuration. Hyperparameter sensitivity analysis further identifies an effective parameter configuration that provides a favorable balance between optimization quality and computational efficiency. Comparative experiments with five representative multi-objective optimization algorithms further verify the superiority and robustness of the proposed method.