| With the development of 5G technology,the number of connected devices has grown exponentially,and emerging real-time applications require high-speed data transmission capabilities.Due to limited wireless spectrum and transmit power,traditional networks struggle to meet the real-time requirements of users.Cloud radio access networks(CRANs)can effectively support real-time performance in wireless communication systems.In CRAN,the efficient utilization of baseband unit(BBU)resources is crucial for ensuring system real-time capabilities.Existing research has not utilized the concept of a centralized BBU pool resources based on traffic load prediction.Additionally,existing deep learning methods incur high optimization costs and have the risk of compromising user privacy as they rely on user data for experiments.Cognitive radio(CR)is an effective technology for improving spectrum utilization efficiency.However,individual CR cannot accurately detect the activity state of primary users(PUs).Cooperative spectrum sensing can enhance detection performance,but it introduces malicious users,thereby compromising system security.Furthermore,as the number of users increases and with limitations on spectrum and energy resources,malicious users will compete for spectrum with PUs,exacerbating the problem of spectrum scarcity.Traditional methods based on received signal strength are ineffective in countering attacks from malicious users as they can mimic the signal attributes and transmission power of PUs.Therefore,targeted research is needed to address the aforementioned issues.In this research,we first proposed an efficient radio resource resources allocation scheme to boost user throughput,using regularized particle filter(RPF).Then we proposed a predictive BnL method to map multiple remote radio heads to one BBU.Lastly we proposed a scheme that detects malicious users present in the network.We divided the detection problem into two:First we presented a localization method called modified adaptive orthogonal matching pursuit(MAOMP)to detect malicious users.Then we proposed a fast,accurate and efficient approach to detect malicious users present in the network using RPF MAOMP.The contributions are as follows:(1)To address the issue of low utilization of radio resources,we proposed an efficient wireless resource allocation scheme that employs an RPF with optimized kernel size and kernel bandwidth.To estimate the next state of the channel,we utilize a Monte Carlo Markov Chain(MCMC)move step based on the Metropolis algorithm.The channel state is computed using maximum likelihood estimation from the temporal channel correlations.By considering the posterior probability and normalized weights,we predict the throughput of the next user state.The main objective of this proposed method is to observe user throughput.Thus,we formulate the optimization problem as maximizing of throughput given the channel state.Solving this non-convex optimization problem is achieved using a branch-and-bound method.The simulation results demonstrate a significant improvement in average throughput,ranging from 5.94%to 52.98%,compared to the Sequential Importance Sampling with Resampling(SISR)method.(2)To tackle the challenge of inefficient utilization of BBU pool resources,we propose a novel Predictive Borrow and Lend(BnL)algorithm based on channel prediction,that facilitates the mapping of BBUs and RRHs.The algorithm leverages an RPF to predict the channel state and intelligently maps multiple RRHs to a single BBU based on its capacity.The primary objective is to maximize the utilization of BBU by effectively aggregating multiple RRHs and map them to a single BBU.In cases where the utilization of a BBU surpasses the upper limit,the algorithm dynamically switches RRHs to alternative BBUs with lower resource utilization.This approach not only enhances efficiency but also reduces latency.In scenarios with a significant number of users,the Predictive BnL method exhibits a noteworthy improvement of 12.80%in both spectral and energy efficiency.Additionally,in scenarios with a smaller user count,our proposed algorithm demonstrates a remarkable spectral and energy efficiency enhancement of up to 37.91%when compared to BnL method.(3)To tackle the challenge of accurately detecting the activity state of primary users for individual cognitive radio users,this study introduces cooperative spectrum sensing.Moreover,to address security concerns arising from malicious attacks in cooperative spectrum sensing,such as primary user emulation attacks(PUEAs),a MAOMP algorithm is proposed.This algorithm aims to detect the presence of PUEAs in CR networks by leveraging user locations.Performance evaluation is conducted using the restricted isometry property.Simulation results reveal that the MAOMP method tends to select base stations with greater location errors between primary users and cognitive users.In comparison to the conventional adaptive orthogonal matching pursuit(AOMP)method,the proposed algorithm exhibits larger residual errors.(4)To address the problem of malicious users imitating legitimate users and competing for spectrum,resulting in spectrum scarcity,an RPF MAOMPbased localization algorithm is proposed to detect the presence of malicious users in the network.To provide a detailed explanation of the algorithm’s convergence speed under real-time conditions,an RPF algorithm is employed to evaluate the performance of the MAOMP.Performance evaluation is conducted using the restricted isometry property.Simulation results demonstrate that the proposed algorithm outperforms existing methods in terms of computational complexity.Furthermore,when the RPF algorithm is utilized,the proposed method achieves a significant improvement in spectral efficiency and energy efficiency of 89.6-90.2%and 85-87%,respectively,compared to OMP,AOMP,and SAMP methods.Additionally,when the RPF algorithm is employed,the spectral efficiency and energy efficiency of the MAOMP method improve by 46.6-76.5%and 83-88.3%,respectively. |
PDF Full Text Request