MATRIX Academic International Online Journal Of Engineering And Technology

An Intelligent Trust-Based Dissemination Framework for VANETs Using Reinforcement Learning and Edge-Assisted Validation

Anurag Gupta — 2026-10-29

Vehicular ad hoc networks are a key communication component
of intelligent transportation systems, where timely and reliable
dissemination of critical information is essential for safety and traffic efficiency.
Nevertheless, rapid topology changes, intermittent connectivity,
and malicious node behavior can degrade dissemination reliability and
increase communication delay. This paper presents a trust-based dissemination
framework that combines context-aware trust evaluation, reinforcement
learning-based dissemination control, and edge-assisted validation
to improve secure and efficient data exchange among vehicular
nodes. Local trust is computed from forwarding reliability, message consistency,
plausibility assessment, historical cooperation, and delay-aware
behavior. Based on these trust observations and network conditions,
a Q-learning agent adaptively selects forwarding, cautious forwarding,
rerouting, holding, or isolation actions. To further enhance decision accuracy
and reduce false positives, nearby roadside edge units perform
regional corroboration and suspicious-event validation for uncertain or
suspicious events. The proposed framework is intended to enhance dissemination
reliability, reduce end-to-end delay, and improve resilience
against malicious behavior in dynamic vehicular environments. Overall,
it offers a scalable and intelligent approach for secure, reliable, and
low-latency vehicular communication.

Algorithm for Parametric Accounting of the Underlying Surface Effect on Radio Wave Propagation Loss Based on Experimental Data

A.A. Kadirov — 2026-10-29

The accuracy of radio wave propagation loss estimation
is limited by the fact that empirical models describe the environment
through an averaged terrain category and do not distinguish
open terrain, vegetation cover and water surface. The transition from
heterogeneous field measurements to a set of parameters suitable for
further modelling remains insufficiently formalized. The aim of this
work is the development and experimental-statistical substantiation of
an algorithm for parametric accounting of the underlying surface type
effect on radio wave propagation loss based on field measurement data.
Primary numerical series of signal level measurements in the 900 and
1800MHz bands along seven paths were used (open terrain: 175 and
130 points; rural single-storey built-up area: 215, 108 and 78 points;
fresh-water surface: 103 and 94 points), together with arrays of 893 and
359 measurements in vegetation cover. The algorithm comprises data
pre-processing, classification of paths by the type of underlying surface,
statistical processing of the samples with regression approximation and
residual calculation, determination of quantitative propagation parameters,
testing of their statistical significance with possible refinement,
and formation of the output parameter set. It was established that the
path loss exponent equals 2.32 and 2.33 for open terrain, 1.45–2.05 for
single-storey built-up areas and 3.22 and 3.23 for the fresh-water surface,
with a standard deviation of residuals of 3.25–9.50 dB. The result of the
algorithm is a parameter set differentiated by surface type and suitable
for the subsequent construction of a surface-conditioned loss correction.

6G Technologies for Vehicular Ad Hoc Networks: A Survey of Enablers, Architectures, and Open Challenges

Dharm Raj — 2026-10-29

Vehicular ad hoc networks (VANETs) are evolving from best-effort safety-message broadcasting toward a data-intensive substrate for cooperative automated driving, remote driving, and immersive in-vehicle services. These applications demand sub-millisecond latency, reliability beyond 1−10−7, centimetre-level positioning, and multi-Gbps throughput under extreme mobility, which exceeds what DSRC, LTEV2X, and 5G NR-V2X can jointly guarantee. This survey reviews the sixth-generation (6G) technologies most relevant to VANETs: terahertz and sub-THz communications, reconfigurable intelligent surfaces, integrated sensing and communication, native artificial intelligence, edge and fog computing with digital twins, integrated space–air–ground networks, and next-generation security and privacy mechanisms. For each enabler we summarise the operating principle, its specific benefit for vehicular links, representative research results, and the vehicular-specific obstacles that remain. We then propose a layered reference architecture for 6G-enabled VANETs, compare the enablers against key performance indicators, and identify open research directions, including mobilityaware beam management at THz, real-time RIS control, trustworthy and explainable AI, and standardisation of IMT-2030 V2X.

Hybrid Deep Learning Model for Skin Cancer Detection Based on CNN and LSTM

Rashmi Sinha — 2026-10-29

Skin cancer is among the most frequently diagnosed malignancies
worldwide, and its outcome depends strongly on how early
melanoma and related lesions are recognised. Convolutional neural networks
(CNNs) have reached specialist-level accuracy on dermoscopic images,
yet they treat each image as a single static input and say little about
how local lesion structures relate to one another. Hybrid models that
couple a CNN feature extractor with a long short-term memory (LSTM)
network have therefore attracted growing interest, because the recurrent
stage can model dependencies across image patches, feature-map regions,
multiple scales or successive examinations of the same lesion. This survey
reviews CNN–LSTM hybrid architectures for skin cancer detection
and classification published between 2017 and 2026. It summarises the
clinical background and public dermoscopy benchmarks, explains the
computational principles of the CNN and LSTM components, and proposes
a taxonomy of hybrid designs according to how the input sequence
is constructed and how the recurrent unit is configured (unidirectional,
bidirectional or attention-augmented). Representative studies are compared
in terms of backbone, recurrent unit, dataset and reported performance,
and the comparison is used to identify recurring weaknesses:
class imbalance, limited diversity of skin tones, weak justification of the
“sequence” in single-image settings, inconsistent reporting and limited
interpretability. The review concludes with research directions, including
longitudinal lesion monitoring, transformer–recurrent hybrids, multimodal
fusion with patient metadata, federated training and lightweight
deployment at the point of care.

Multi-Model Forensic Ensemble and Partially Fine-Tuned CLIP Detector for Identifying AI-Generated Images: The Xabarnavis System Case Study

Solijonov Abduxoliq — 2026-10-29

Photorealistic images generated by generative models (Midjourney, DALL·E, Flux, Stable Diffusion, GPT-image) are increasingly becoming a means of spreading disinformation in the mass-media information flow. This paper proposes a two-stage approach developed for the Xabarnavis media-forensics platform: (i) a degradation-robust binary detector in which the last 8 blocks of the CLIP ViT-L/14 encoder are fine-tuned, and (ii) an ensemble that performs weighted fusion of the outputs of 16 independent detectors. The model was trained on 134,108 images from the open-source OpenFake and CommunityForensics datasets (including 14,404 images from commercial generators) and evaluated on four held-out test sets: AUC 0.968 on new generators, AUC 0.958 on social-media images, and AUC 0.984 and 0.998 on the external GenImage++ and DRAGON datasets, respectively. The mathematical formulation of the task, loss function, temperature calibration, ensemble fusion formula, and model scheduling under constrained GPU memory are described.

Prediction of Groundwater Changes Using Artificial Intelligence Models Based on Remote Sensing Images of Surface Water Bodies

Dadajonova Zilola Botirjon kizi — 2026-10-29

The study analyzes the annual changes in surface water bodies based on multi-platform remote sensing satellite imagery and links these changes to the depletion of groundwater resources. Among transformer-based architectures, the SegFormer model was selected and trained using water body images from 2020 to 2025. The model identified a linear decreasing trend in water surface area and predicted a decline in groundwater reserves for 2026. According to the results, the water-covered area decreased by 19.6% during 2020–2025 and is expected to decline by 23.9% by 2026, with an average annual reduction rate of approximately 3.9%.