The problem of spectrum scarcity has raised the prospect of coexisting network. For example, LTE-LAA & Wi-Fi can operate in the same frequency band. However, interference between the two different radio access technologies can severely reduce the performances of both networks. To mitigate this interference, the concept of an almost blank subframe (ABS) has been proposed. However, because the LTE-LAA macrocell must remain silent, this provides no incentive for its throughput enhancement. In this paper, we propose a transaction between LTE-LAA and Wi-Fi using ABS together with offloading. We model a time-resource utility model by jointly considering the resource consumed for ABS and offloading, then formulate a coalition game. To achieve the solution of a coalition game, we consider two fairness concepts: min-max and Shapley fairness. We then verify the performance of the proposed game using simulations. It is shown that the proposed game model can improve resource utilization through the joint use of ABS and offloading.

This paper proposes a method for gradually calibrating channel non-reciprocity in an ultra-dense cell-free massive multiple-input multiple-output (MIMO) system. Many studies assume perfect channel reciprocity in time division duplex (TDD) systems. However, reciprocity is not valid in actual channels. Therefore, determining the reciprocity-calibrated precoding matrix is important for the performance of the network. In a cell-free massive MIMO system, user equipment (UE) is served by many access nodes (ANs). Since there are many nodes, it is difficult to calibrate the matrix simultaneously. Using the proposed method, the network sequentially calibrates the precoding matrix in order of influence. Thus, the proposed method can obtain a large gain compared to a system without calibration by performing only minimal channel calibration. To show the capacity gain of the proposed method, a simulation was performed considering the overhead required for channel calibration.

Machine learning algorithms have been extensively explored for the feedback of multiple-input multiple-output (MIMO) channel state information (CSI) in orthogonal frequency division multiplexing (OFDM) systems. However, their viability in sixth-generation (6G) wireless communication systems, operating in the terahertz (THz) band, remains uncertain. To address this, we propose ChannelTransformer, a transformer-model-based CSI feedback scheme that incorporates multi-head self-attention and a CSI-feedback-aware transformer structure, and a lightweight user equipment(UE) model. Through simulations in the DeepMIMO O1 scenario at 140GHz, ChannelTransformer demonstrates superior performance in terms of normalized mean square error (NMSE) and cosine similarity across various feedback lengths compared to conventional schemes with a much smaller UE model size.

In this work, we present the innovative Concrete Feedback Layers, designed to enable genuine bit-level, end-to-end Channel State Information (CSI) feedback using deep learning techniques. Overcoming the limitations of traditional discrete operations that impede gradient flow, these layers leverage the concrete distribution to facilitate efficient learning processes. Our extensive simulations reveal that these layers significantly enhance digital CSI feedback, achieving superior performance in terms of Normalized Mean Squared Error (NMSE) and cosine similarity compared to conventional feedback models. Furthermore, the integration of the Concrete Feedback Layers with the Feedback Bit Masking Unit (FBMU) allows for authentic bit-level variable-length CSI feedback, while maintaining a single adaptable model for various feedback lengths. This advancement marks a major leap forward in deep learning-based CSI feedback methods, potentially revolutionizing 6G communication systems with its flexibility and efficiency.

We propose an adversarial attack for machine-learning-based network intrusion detection systems that selectively alters only the most influential features. Unlike conventional attacks such as the Fast Gradient Sign Method (FGSM) and multi-step Projected Gradient Descent (PGD), which can perturb all feature dimensions, our DeepSHAP-guided masked-PGD approach focuses its updates on a small set of highly influential features while still achieving competitive success rates on the X-IIoTID dataset. Experimental results indicate that it yields lower autoencoder-based reconstruction error and uses much sparser perturbations, making its adversarial examples harder to detect. By systematically varying the perturbation budget and the number of targeted features, we show the trade-off between stealth and effectiveness. These findings demonstrate the difficulty of detecting adversarial examples guided by feature importance.