With the advent of the information age, the effective identification of sensitive information and the leakage of sensitive information during the transmission process are becoming increasingly serious issues. We designed a sensitive information recognition and encryption transmission system based on a decision tree. By training sensitive data to build a decision tree, unknown data can be classified and identified. The identified sensitive information can be marked and encrypted to achieve intelligent recognition and protection of sensitive information. This lays the foundation for the development of an information recognition and encryption transmission system.

As the number of layers of deep learning models increases, the number of parameters and computation increases, making it difficult to deploy on edge devices. Pruning has the potential to significantly reduce the number of parameters and computations in a deep learning model. Existing pruning methods frequently require a specific distribution of network parameters to achieve good results when measuring filter importance. As a result, a feature map similarity score-based pruning method is proposed. We calculate the similarity score of each feature map to measure the importance of the filter and guide filter pruning using the similarity between the filter output feature maps to measure the redundancy of the corresponding filter. Pruning experiments on ResNet-56 and ResNet-110 networks on Cifar-10 datasets can compress the model by more than 70% while maintaining a higher compression ratio and accuracy than traditional methods.

Spread Spectrum (SPSP) Communication is the theoretical basis of Direct Sequence Spread Spectrum (DSSS) transceiver technology. Spreading code, modulation, demodulation, carrier synchronization and code synchronization in SPSP communications are the core parts of DSSS transceivers. This paper focuses on the code synchronization problem in SPSP communications. A novel code synchronization algorithm based on segment correlation is proposed. The proposed algorithm can effectively deal with the informational misjudgment caused by the unreasonable data acquisition times. This misjudgment may lead to an inability of DSSS receivers to restore transmitted signals. Simulation results show the feasibility of a DSSS transceiver design based on the proposed code synchronization algorithm. Finally, the communication functions of the DSSS transceiver based on the proposed code synchronization algorithm are implemented on Field Programmable Gate Array (FPGA).

SM9 algorithm plays a very important role in the field of information security in our country, it can effectively solve the problem of certificate management of PKI. The dot multiplication operation in the SM9 algorithm takes up a lot of computing time, which seriously affects the operation efficiency. The traditional modular multiplication algorithm uses a simple shift and addition method. To shorten the time consumed by the operation, this paper proposes an improved modular multiplication algorithm that supports a four-stage pipeline. In addition, this paper also designs and improves algorithms such as point addition, point doubling, and point multiplication. Through FPGA simulation and verification, it only takes 0.848ms to realize one point multiplication operation, and the occupied resources are 18526 look-up tables (LUTs) and 13982 flip-flops (FFs). This research has great significance for the development of SM9, making it have wider application value.

Deep learning techniques are frequently utilized and produce effective results in the classification of encrypted traffic. In the current encryption traffic classification process, the network traffic characteristics are not sufficiently extracted, which is a concern. An encrypted traffic classification model based on raw network traffic and its spatiotemporal characteristics is proposed in this paper. The raw network traffic is divided into sessions, and the packets inside each session are then split into 784-byte slices, and the traffic is then described using the slice data. The time feature vector and the spatial feature vector are then created by combining ResNet and GRU models to generate features from raw network data in parallel. The traffic is then classified using the combined features. According to experimental findings, the proposed model's recognition accuracy on the ISCX-NonVPN-VPN2016 dataset reached 99.36%, which is an improvement over other approaches currently in use.