With the rapid development of artificial intelligence technologies, the demand for training data has surged, exacerbating risks of data leakage. Despite increasing incidents and costs associated with such leaks, data leakage prevention (DLP) technologies lag behind evolving evasion techniques that bypass existing sensitive information recognition (SIR) models. Current datasets lack comprehensive coverage of these adversarial transformations, limiting the evaluation of robust SIR systems. To address this gap, we introduce DataSIR, a benchmark dataset specifically designed to evaluate SIR models on sensitive data subjected to diverse format transformations. We curate 26 sensitive data categories based on multiple international regulations, and collect 131,890 original samples correspondingly. Through empirical analysis of real-world evasion tactics, we implement 21 format transformation methods, which are applied to the original samples, expanding the dataset to 1,647,501 samples to simulate adversarial scenarios. We evaluated DataSIR using four traditional NLP models and four large language models (LLMs). For LLMs, we design structured prompts with varying degrees of contextual hints to assess the impact of prior knowledge on recognition accuracy. These evaluations demonstrate that our dataset effectively differentiates the performance of various SIR algorithms. Combined with its rich category and format diversity, the dataset can serve as a benchmark for evaluating related models and help develop future more advanced SIR models. Our dataset and experimental code are publicly available at https://www.kaggle.com/datasets/fanmo1/datasir and https://github.com/Fan-Mo-ZJU/DataSIR.

Spectrum resource allocation and flow control are important considerations in dual-connectivity (DC) technology. Currently, limited research exists on spectrum resource allocation and flow control when user equipment (UE) employs DC mode within the cloud radio access network (C-RAN) architecture. Therefore, we propose a spectrum resource allocation scheme for DC mode in C-RAN architecture. The scheme formulates the problem as an optimization to maximize a utility function, allocating spectrum resources based on an operator’s revenue, UE’s service priority, signal-to-interference-plus-noise ratio (SINR) at the UE’s channel, and system throughput. We also proposed a flow control method for various application scenarios in 6G wireless networks. The method selects transmission path for downlink data based on delay estimation and the channel’s SINR. Simulation results demonstrate the proposed scheme’s effectiveness in improving frequency resource utilization, operator’s revenue, and system throughput compared to existing schemes, while significantly reducing bit error rate for enhanced ultra-reliable low latency communication with minimal transmission delay increase.

Distributed base station caching is an efficient way to reduce the network traffic load in advanced cellular networks, but suffers from a significant network delay due to the multi-hop transmission, leading to a poor quality of service. The presence of a few shortcut direct links transforms a regular network to a small world network resulting in achieving properties of high clustering coefficient and low average path length. In this paper, we propose a delay-optimized small world model for base station caching network to reduce the average network delay by decreasing multi-hop distance. In particular, we first study the degree-based and closeness-centrality shortcut addition strategies. To enhance these two strategies, we propose the delay-optimized shortcut addition strategy to decrease average network delay. The numerical results show that the proposed delay-optimized small world model can reduce the average network delay under remaining the clustering coefficient. Meanwhile, the average network delay with the proposal is 31.26% lower than that with the degree-based strategy, and is 18.78% lower than that with the closeness-centrality strategy.

The performance analysis under various precoding strategies in heterogeneous cloud radio access networks (H-CRANs) is increasingly highlighted. In this paper, we focus on suppressing the inter-tier interference between macro base stations (MBSs) and remote radio heads (RRHs), which is severely affected by the application of precoding schemes. Two precoding schemes including interference cancelation (IC) and beamforming (BF) are characterized for performance evaluation. The average bit error rate (BER) and system sum capacity under these two schemes are derived with closed-form expressions. Monte Carlo simulations are performed to verify our analytical results and demonstrate the effectiveness of precoding schemes. Furthermore, we explore this problem analytically and demonstrate that BF is preferred when the signal-to-noise ratio (SNR) of MBSs is relatively low and IC outperforms BF at the high SNR region.

Cloud radio access networks (C-RANs) have attracted considerable attention because of the capability of meeting the exponential increasing traffic demand in the future communication systems. In this paper, we consider the segment training based channel estimation in C-RANs. As the classical minimum mean-square-error estimator has cubic complexity in the dimension of the covariance matrices, due to the inversion operation, we propose a low-complexity channel estimator by means of the \emph{L}-degree matrix polynomial expansion, which can significantly reduce the computational complexity without degrading much performance. The numerical results are presented to verify the proposed channel estimators, and the simulation results show there are significant performance gains from our proposal.