Cognitive radio networks (CRNs) are highly vulnerable to spectrum spoofing, PU emulation (PUE), and harmonic RF manipulation attacks, which significantly degrade dynamic spectrum access reliability and communication security. This paper proposes a harmonic trust spectrum modeling framework for cognitive radio intrusion detection using multi-domain RF intelligence and adaptive machine learning. The proposed system integrates harmonic spectrum analysis, trust oscillation modeling, spectral entropy analysis, wavelet decomposition, and power spectral density (PSD) characterization to detect anomalous RF behaviors. Three heterogeneous RF datasets were used for experimental evaluation: the RadioML 2016.10A modulation dataset, the Oracle RF fingerprinting IQ dataset, normal signals generated by GNU Radio, and malicious PUE RF signals. The proposed methodology extracts advanced harmonic trust features including spectral entropy, spectral flatness, harmonic trust instability index (HTII), and oscillatory trust divergence (OTD), which are subsequently processed using an adaptive XGBoost-based intrusion detection system. Experimental results demonstrate that malicious RF users exhibit significantly higher spectral instability, irregular harmonic distributions, elevated entropy behavior, and oscillatory trust divergence, compared to legitimate users. Furthermore, studies covering fast Fourier transform (FFT) harmonic spectra, PSD analysis, wavelet coefficient analysis, trust oscillation curves, and confusion matrix evaluations confirm the effectiveness of the proposed framework. The proposed harmonic trust spectrum model provides a computationally efficient and scalable solution for next-generation secure cognitive radio communications and RF cyber defense systems.
The paper considers the adaptive forward error correction (FEC) technique for real-time video transmission over a feedback-free channel with limited bandwidth and non-stationary interference. A hybrid Reed-Solomon coding method is proposed in GF(2\textsuperscript{8}), combining differentiated redundancy across specific frame types (I, P, and B), with adaptive code parameter control driven by physical layer channel state estimation. The channel is modeled by a three-state Markov process called Gilbert-Elliott-Jamming (GEJ), in which an explicit jamming state is added to the classical Gilbert-Elliott model. A closed-form expression for the block decoding failure probability under the erasure model is derived, and applicability bounds for representative RS configurations are determined. A redundancy adaptation function with a guaranteed upper bound on overhead is introduced together with a control algorithm. Experiments conducted under four scenarios (stable Wi-Fi, mobile 4G, tactical channel and active jamming) show that the mission success rate increases from 60% for fixed RS and 50% for SRT to 78% under sustained jamming, while the time to first frame decreases from 600 - 1200 ms to 200 ms. The method is implementable on the ESP32-S3 and ARM Cortex-M4 platforms with an encoding throughput of 4 to 8 Mbps.
Demand for quick lossless data compression systems has become more important recently due to the growing deployment of Internet of Things (IoT) devices in terrestrial and underwater environments, under restricted bandwidth, latency and energy consumption conditions. Current dictionary-based and entropy-driven compression methods depend on reactive adaptation and fixed block processing, which makes them less useful in situations where sensing is constantly changing. This paper presents a self-optimizing entropy prediction-assisted lossless compression framework (EP-SLZW), where the compression technique is based on the expected data redundancy and current channel condition. To achieve optimal results, lightweight entropy prediction is employed in combination with adaptive block segmentation and dual dictionary learning between processing and compression. The optimization component of the proposed method enhances the compression strategy by taking care of all important environmental factors to provide the best solutions for radio frequency and acoustic communication channels. The method is evaluated with the help of the MQTT and CoAP protocols through NetSim simulations and hardware experiments. The results show that the proposed approach performs better in comparison to conventional LZW as well as Huffman and run-length encoding techniques, providing better throughput, compression ratio, and less end-to-end delay when dealing with resource-constrained IoT devices.
Pattern division multiple access (PDMA) is recognized as a promising non-orthogonal multiple access technique for overloaded wireless systems, capable of being used for multiplexing multiple users over a limited set of resources. However, the real performance of PDMA is determined not only by the access principle itself, but also by the joint interaction between pattern design, transmit power allocation, and receiver interference cancellation. This paper proposes a fairness PDMA scheme for overloaded downlink systems based on joint pattern assignment, power allocation, and adaptive successive interference cancellation (SIC). The design aims to improve spectral efficiency and user fairness under real residual-interference conditions. Simulation results show that the proposed PDMA consistently outperforms orthogonal multiple access (OMA) and fixed-pattern PDMA techniques. At 30 dB, the proposed scheme achieves an average sum rate of approximately 14.5 bit/s/Hz under ideal SIC, compared with nearly 12 bit/s/Hz for OMA and approx. 8.5 bit/s/Hz for fixed-pattern PDMA. In terms of fairness, at an overload factor of \textlambda = 1.5, the proposed method attains a Jain's fairness index of approx. 0.84, whereas OMA and fixed-pattern PDMA achieve nearly 0.58 and 0.44, respectively. These results confirm that an adaptive joint design allows to obtain both high throughput and balanced user performance in overloaded PDMA systems.
Low-power acoustic telemetry remains a significant challenge in Internet of Things monitoring systems deployed in remote environments. This paper presents an embedded fast Fourier transform framework for communication-efficient acoustic monitoring of honey bee colonies. Instead of transmitting raw audio streams, the proposed approach extracts compact spectral descriptors directly on an embedded sensing node and transmits only a small feature vector using a low-power cellular network. The framework is evaluated using labeled queenright and queenless colony recordings. The proposed solution targets resource-constrained ESP32 class IoT nodes operating over LTE-M and NB-IoT networks. The analysis covered such parameters as dominant frequency, peak amplitude, mean spectral amplitude, spectral centroid, spectral entropy, and band energy extracted from the 200 - 400 Hz band. The results showed that dominant frequency alone did not significantly differentiate colony states at the file level, while mean spectral amplitude remained statistically significant. Queenless recordings also exhibited higher dominant frequency variability. The proposed approach reduces the transmitted payload by more than three orders of magnitude while remaining compatible with resource-constrained ESP32 class IoT devices. The results demonstrate that the extraction of embedded acoustic features is a practical method for scalable smart beehive monitoring under strict memory, power, and bandwidth constraints.
Automatic modulation classification (AMC) for 5G-Advanced and 6G networks must blindly identify waveforms from received signals under realistic channel impairments, enabling cognitive radio dynamic spectrum access and interference avoidance. No prior work has simultaneously applied machine learning to classify all eight leading waveforms (UFMC, GFDM, FBMC, NOMA, OFDM-IM, OTFS, ODDM, and AFDM) under realistic channel impairments, nor quantified the minimum feature set for resource-constrained deployment. We present a framework that (i) extracts a 38-dimensional feature vector that includes three novel channel-aware characteristics (amplitude fading variance, phase discontinuity, and frequency drift); (ii) benchmarks nine machine learning classifiers, including an FC-MLP deep learning baseline and five feature selection methods, on 201600 signals across twelve channel conditions (nine custom plus three 3GPP TDL profiles) and seven SNR levels, with leakage-free feature selection; and (iii) identifies a compact 10-feature subset validated with Bonferroni-corrected McNemar tests and Wilson confidence intervals. FC-MLP achieves 99.09% accuracy; ensemble-boost (99.04%) and random forest (99.02%) are statistically equivalent. The 10-feature random forest reaches a score of 98.90% within 0.12 pp of the full feature baseline at a cost that is 74% lower and with a 0.071 ms inference per block. The five-fold cross-validation confirms stability (98.54%, Wilson 95% CI: 98.49%, 98.59%). Per channel accuracy ranges from 98.87% (Rayleigh) to 99.98% (AWGN/Rician); 3GPP TDL-A/B/C profiles confirm transferability to 5G NR. The three channel-aware features yield up to 3.1% gain under double-selective fading and an average overall improvement.
This paper investigates the performance of reconfigurable intelligent surface (RIS)-assisted LoRa networks. Specifically, we consider a LoRa system enhanced by RIS under the influence of hardware impairments and asymmetric channel conditions. A closed-form expression for the outage probability at end devices is derived using the method of moments. The accuracy of the proposed analytical framework is extensively validated through Monte Carlo simulations. Several important insights are drawn from both the theoretical analysis and simulation results. In particular, the system's performance is significantly enhanced by an increase in the number of RIS elements and the transmission power of the gateway. Furthermore, comparisons with related works described in the literature are made to show that the proposed system outperforms these existing approaches simply by increasing the number of RIS elements. Additionally, we reveal that a higher spreading factor (SF) does not necessarily lead to worse performance than a lower SF, and the impact of hardware impairments is found to be minor under typical operating conditions.
Multi-controller SDN environments suffer from a blind spot when it comes to detecting low-rate DDoS attacks. Each controller sees only its own traffic slice, meaning that an LDDoS campaign looking, at every controller, like background noise is still capable of draining the network. Federated learning (FL) is a reasonable answer to this challenge, due to such controllers sharing model updates rather than raw logs. However, the published literature on FL-based detection is fragmented enough that the results have not been systematically compared up to date. We analyze 39 papers published between 2020 and 2026. 35 of those reported quantitative results, with the pooled mean detection precision equaling 98.25% (SD ±0.91) and the mean F1 score amounting to 97.98% (SD ±1.10). Federated models averaged an accuracy score of 98.33%, compared to 98.06% for centralized approaches - a 0.27 pp gap that is practically negligible. LSTM and hybrid CNN + RNN architectures ranked the highest in terms of the most metrics. Four aggregation strategies were mentioned repeatedly: weighted aggregation, asynchronous FL, personalized FL, and standard FedAvg. The widest gap we identified was in the datasets. No available benchmark simultaneously models multi-controller SDN topology, low-rate attack patterns, and heterogeneous traffic distributions across various controllers. Until that changes, high-accuracy scores on CICIDS2017 or CICDDoS2019 should be interpreted with some caution.
The CR technology enhances spectrum utilization by allowing access to unused licensed channels, while spectrum sensing allows secondary users to verify channel availability before the transmission. This study relies on the LoRaCog framework, a solution integrating the CR technology with LoRa LPWAN networks, to evaluate the performance of eigenvalue-based detection algorithms, such as maximum eigenvalue detection (MED), maximum to minimum eigenvalue (MME), energy-to-minimum eigenvalue (EME) and maximum-to-mean eigenvalue detection (MMED), with the comparisons based on energy detection (ED). The said algorithms were evaluated under three scenarios characterized by an increasing degree of complexity. These included the following: an ideal additive white Gaussian noise (AWGN) channel, followed by a multipath fading channel with noise uncertainty using a SISO receiver and, finally, a SIMO multiantenna receiver system. The simulation results for the AWGN channel showed that the ED algorithm achieved the best detection probability and the lowest sensing time. When multipath fading and noise uncertainty were introduced, eigenvalue-based algorithms achieved higher detection probabilities while maintaining comparable detection times. The MME algorithm achieved the highest detection probability when used with the SIMO multi-antenna reception system.
In clustered wireless sensor networks (WSNs), re-shaping the topology can redistribute cluster head load, but each such task consumes energy. This paper studies the refresh timing problem in static clustered WSNs, where the controller decides not only whether to rebuild the topology but also determines the time over which the selected topology remains active. The proposed method formulates topology maintenance as a semi-Markov adaptive holding-time control problem. At each control epoch, the controller selects a refresh indicator, a target cluster count, and a holding time. The topology builder uses explicit cluster head election, nearest head member association, and intra-cluster chain forwarding with one-hop cluster head transmission to the base station. Under nominal deployment, the proposed controller reaches a half-node death (HND) point of 1969.1 ±8.4 rounds with 0.104 J of control energy, while periodic refresh with T = 10 reaches 1819.7 ±32.6 rounds and consumes 1.133 J. Across seven tested deployment scenarios, the proposed method gives a higher HND point with lower control energy than the tested refresh-enabled baselines. Therefore, the method is positioned as a lifetime overhead control mechanism, favoring lower control energy and longer mid-life operation, whereas periodic refresh remains preferable when delivery performance is the primary objective.
This paper presents the network architecture and empirical performance analysis of the Proof of Concept (POC) for a stateless Tor-based communication system designed for privileged communication. Unlike existing secure messaging platforms relying on centralized server infrastructures, persistent session states, or identifiable network endpoints, the proposed solution achieves server-side and client anonymity simultaneously through the integration of Tor hidden services v3, stateless application design, and containerized microservice decomposition. We formally describe the system's model and its constituent components: an application server, an ephemeral identity registry, and a browser-based client operating over WebCrypto. Next, we analyze performance of the network layer across 100 measurement cycles. Empirical results confirm that cryptographic operations contribute less than 2 ms of overhead relative to dominant Tor circuit latency (mean value of 8100 ms per circuit). Immunity to traffic, session linkability, and server deanonymization are examined against a realistic network adversary model. POC is compared to SecureDrop, Ricochet, and Signal in terms of five architectural properties and is shown to be the only system under evaluation satisfying all five requirements simultaneously. Deployment considerations for production-grade privileged communication environments, including operational security procedures for public key registration, are discussed as well.
Nowadays, social media impact all aspects of our lives, making us vulnerable to fraud and scams. Bots are believed to be the most prevalent form of malware that may be found in social media environments. New detection methods are required to keep up with the pace of their continuous advancement. This paper offers an overview of machine learning-based bot detection methods. The study revealed that the effectiveness of machine learning (ML) models can be significantly hindered by redundant and irrelevant features present in the datasets, which can lead to performance degradation. A hybrid feature selection (FS) combining characteristics of the genetic algorithm (GA) and the mutual information (MI) approach is proposed to overcome this challenge. The proposed method is evaluated using the following approaches: random forest (RF), decision tree (DT), support vector machine (SVM), and logistic regression (LR). Compared to the state-of-the-art models, the proposed method is capable of efficiently identifying bots using only a small number of features. For the dataset used, we achieved a classification accuracy of 0.99 using 4 features only.
The message queuing telemetry transport (MQTT) protocol is widely adopted in smart home IoT ecosystems despite its default configuration failing to offer adequate protection against eavesdropping or payload manipulation. This study addresses an important research gap and attempts to determine whether AES-128 payload encryption is capable of securing MQTT transmissions without degrading the effectiveness of machine learning-based intrusion detection systems (IDS). Three security configurations, namely TLS, payload encryption, and token-based authentication, deployed on the ESP32 microcontroller family, are compared and their impact on message latency is measured. Experimental results show that the AES-128 encryption overhead remains at below 25% of the message publication time on ESP32-S3. To evaluate the robustness of IDS under encryption, we apply a reproducible modification to the MQTTset benchmark dataset that replaces variable-length plaintext payloads with fixed-length ciphertext representations while simultaneously preserving feature semantics and labeling consistency. 5 out of 6 evaluated classifiers maintained their accuracy level at above 99% on the modified dataset, with tree-based and neural models showing no meaningful degradation. Only Naive Bayes proved unsuitable, with its accuracy dropping from 98.79% to 62.15% due to its independence assumptions being violated by cryptographic uniformity. These results confirm that AES-based MQTT payload encryption is a practical and efficient security measure for resource-constrained IoT environments, provided that appropriate classifiers are employed.
The aim of this study is to model the impact of main reflector deformations in a double-reflector spherical antenna system on the phase distribution of the electromagnetic field across the aperture and the associated gain loss. The study focuses on the antenna of the ROT-54/2.6 radio-optical telescope (Herouni radio telescope) - a spherical double-reflector system with a fixed primary reflector with a 54 m diameter, composed of 3738 panels. An analytical model is developed to evaluate phase distortions induced by deviations from the spherical geometry. The model computes local phase shifts across the aperture and predicts gain degradation using Ruze's formula which relates the RMS surface error to efficiency losses. This approach is important for pre-alignment procedures and functional restoration of the antenna, enabling geometry corrections prior to full-scale observations. Based on terrestrial laser scanning (TLS) data, the methodology allows for a quantitative assessment of structural phase errors and corresponding gain degradation, confirming its suitability for practical diagnostics of large reflector systems.
Thanks to its improved spectral efficiency and immunity to frequency selective fading, OFDM with index modulation (OFDM-IM) has become a perspective option. Unfortunately, OFDM-IM systems are vulnerable to security risks due to their inherent openness encountered in wireless communications. Conventional encryption techniques, which focus on the upper layers, add complexity and might not be enough to fend off malicious attacks. To improve the selection of subcarrier indexes and modulation of data symbol modulation, this work proposes a new chaotic encryption approach for OFDM-IM systems that uses Lorenz chaotic maps. Comprehensive simulations show that, in comparison to traditional methods, the proposed approach provides better security against eavesdropping while maintaining transmission reliability.
Subspace-based direction of arrival (DOA) estimation algorithms, such as MUSIC and ESPRIT, are designed for adaptive smart antenna arrays. However, these subspace methods require a large number of signal snapshots and sufficient angular separation between signals to provide an accurate DOA estimation of RF signal sources. Moreover, their resolution degrades significantly in severe noise scenarios. This study proposes a swarm intelligence (SI) algorithm, known as horse herd optimization (HOA), to address these limitations. An optimizer is employed as a direction-finding method to estimate the directions of arrival (DoAs) of incident signals impinging on a linear array of half-wavelength dipole (HWD) antennas by examining the global minimum of a non-linear cost function. This cost function is defined as the difference between the actual and estimated angles and is used to evaluate candidate solutions. Simulation results of the proposed algorithm have been compared with other recognized algorithms, including ESPRIT, root-MUSIC, and PSO, to verify estimation accuracy, convergence behavior, robustness against the number of elements, noise, and snapshots over Monte Carlo trials. It has been observed that the suggested HOA achieves better performance with a few snapshots, outperforms PSO and subspace-based methods when it comes to estimating DOA of incoming signals, particularly in a low signal-to-noise ratio (SNR) environment, and even when only fewer snapshots are available.
The article is devoted to the analysis and assessment of the efficiency of fiber optic systems whose primary objective is to transmit data. The efficiency of information transmission systems depends on numerous indicators, such as interference immunity, speed, energy efficiency, cost, development time, and design. However, from the user's point of view, quality of service is determined primarily by transmission speed and reliability. This article compares two systems. The first corresponds to the modern paradigm: one user - one transmission channel. In the other, the number of users is provided with a complex channel for transmitting symbols of the alphabet of a certain system of residual classes. At the same time, the transmission speed in the residual class system - compared to the classical multiplexing method - decreases slightly to 28/32, while the reliability (determined based on the probability of failures) increases by several orders of magnitude. The work proves a lemma that determining the optimal alphabets of residual class systems allows to optimally approximate the transmission speed of modules to binary coding systems. An analysis of the non-linear loss function, which considers the speed parameters and probabilistic reliability indicators, is performed as well.
Active intelligent reflecting surfaces (IRS) with phase-shift and amplifier capabilities have arisen as a solution relied upon to improve spectral/energy efficiency of wireless systems, as they outperform conventional passive techniques/without IRS assistance. In this work, the simultaneous wireless information and power transfer (SWIPT) downlink is supported by an active IRS, where a multi-antenna base station (BS) broadcasts both information and power to multiple hybrid power-splitting (PS) users. The target of sum data rate maximization is to study the constraints of user energy harvesting thresholds and power transmission limitations of BS and active IRS. To tackle this complicated issue, iterative algorithms are proposed to find the optimal beamforming vector, PS coefficients, and IRS parameters, as amplification factors and phase shift. A joint optimization framework using alternating optimization, semidefinite relaxation, and non-convex approximations is used. Finally, simulation experiments are performed to assess that the proposed iterative algorithms of the active IRS scheme converge fast and achieve better sum rate results than conventional baseline schemes.
Precise classification of modulation in cooperative relaying networks remains challenging in the presence of carrier frequency offset (CFO) and imperfect channel state information (CSI). This paper conducts a comprehensive comparative analysis of automatic modulation classification (AMC) methods for distributed space-time block-coded orthogonal frequency division multiplexing (DSTBC-OFDM) systems under these impairments. A unified simulation framework is developed that combines pilot-assisted CFO and CSI estimation with higher-order statistics (HOS)-based feature extraction. Four widely used machine learning classifiers, i.e. feedforward neural network, support vector machine, random forest classifier, and adaptive boosting, are benchmarked under identical channel and noise conditions. Monte Carlo simulations are performed across varying SNR levels and fading scenarios, enabling a fair assessment of classification accuracy, robustness to residual estimation errors, and relative computational complexity. The results provide practical insights into the strengths and limitations of each classifier in cooperative STBC-OFDM environments, offering valuable guidelines for selecting AMC techniques in future cooperative wireless systems.
