Iraqi Journal of Information and communication technology

Sparse Code Multiple Access (SCMA) is an extremely effective non-orthogonal multiple access technology
that enables communication between users who have limited orthogonal resources currently. Traditional
SCMA methods use manually designed codebooks, potentially leading to subpar performance owing to inadequate
optimization for certain encoders. A Deep Neural Network (DNN) is used to produce a deep learning for SCMA
codebook using Stochastic Gradient Descent (SGD). This enables the model to determine the optimal weights and
biases for generating precise predictions. The proposed approach surpasses existing techniques in a Rayleigh fading
channel due to its reduced Bit Error Rate (BER), enhanced Minimum Euclidean Distance (MED), and diminished
complexity compared to prior SCMA frameworks

Edge-cloud computing paradigms increase the Quality of Experience (QoE) for real-time applications
by offering many benefits, such as reducing the response time. Many strategies are proposed to handle the data
in the edge cloud environment. Hence, where to execute the data generated by the end device is considered an
important issue. In this paper, an Artificial Intelligence (AI) model is proposed based on a neural network for
workload allocation decisions. The model deals with an AI application that runs on an edge server and a cloud
center. The model was trained using a pre-generated dataset based on several features. The features considered
are the data size, model complexity, application priority, edge server utilization, and delay of execution on the
edge and the cloud. The model decides where to perform the task generated by the end device, either in the edge
server or on the cloud. Four AI applications are considered. The model has been implemented using Tensorflow
platform with the required libraries. The proposed model employee multi-feature with multi-application addressing
workload allocation decisions in a hybrid edge-cloud environment by creating and utilizing a dataset based on
proposed algorithm. The proposed model achieved accuracy reached 98.3% and reduced the response time of task
execution compared to the based line approach considered in this paper.

Long-Range Wide-Area Networks (LoRaWAN) is among the foremost favored techniques for wireless
communication in Internet of Things (IoT) applications, owing to its simplicity and versatility. It utilizes an Adaptive
Data Rate (ADR) approach at both the network server and end device levels. In order to provide the necessary
downlink radio parameters, the network server could use the allocation of a resource strategy to control the transmit
power and spreading factor. Therefore, this work suggests a Uniform Distribution ADR (U-ADR), network servercontrolled
adaptive data rate, to deal with the fast-fluctuating signal-noise ratio of arriving packets at the network
server. The suggested approach seeks the most effective allocation of radio parameters to the end devices to reduce
the total network energy consumption and improve the average packet delivery ratio of the network. The findings
depict that the suggested approach enhances the average Packet Delivery Ratio (PDR) by 102% and 77.94% for
a single and two gateways deployment in an urban scenario respectively while in sub-urban improved by 65.14%
and 33.53%. The total Network Energy Consumption (NEC) was reduced by 9% and 16.55% for single and two
gateways in the urban scenario while in the sub-urban scenario reduced by 15.92% and 12.53%.

Image enhancement in low-light conditions has gained significant attention in recent years due to its
importance in improving visual clarity and uncovering hidden details in poorly illuminated images. This study
focuses on the application of classical methods for enhancing color and brightness, as these approaches provide a
practical balance between efficiency and performance compared to computationally intensive modern techniques.
Classical algorithms were applied, modified, and extended to develop lightweight solutions capable of improving
image quality through relatively simple processes. In this work, a set of classical enhancement methods was tested on
images captured in night conditions. Performance was evaluated using metrics such as the Structural Similarity Index
Measure (SSIM) and the Signal-to-Noise Ratio (SNR), which assess improvements in brightness, color quality, and
detail preservation. Additionally, three novel techniques were proposed: enhanced Hue, Saturation, Value (HSV)
through scaling of the Value channel, Custom HSV-based Brightness and Saturation Scaling, and an Entropybased
Hybrid Enhancement combining HSV and Lab* color spaces. Results showed that the proposed methods
outperformed conventional histogram equalization, with the best-performing approach achieving up to a 41.4%
improvement in Structural Similarity Index Measure. Overall, the findings demonstrate that classical and improved
lightweight methods remain effective and computationally efficient for low-light image enhancement

The issue of sensor coverage in Wireless Sensor Networks (WSNs) is crucial, particularly as these
networks are deployed in military applications for the armed forces as well as in civilian health applications.
Therefore, improving coverage and communication while minimizing interference between sensors is essential. This
paper presents a hybrid meta-heuristic approach to optimizing node deployment in WSNs using a modified Particle
Swarm Optimization (mPSO) and Ant Lion Optimization (ALO) algorithms. The Particle Swarm Optimization
(PSO) algorithm was applied for global search, while the ALO focused on internal search within the Region of
Interest (ROI). Initially, nodes are deployed randomly within the ROI. The algorithm then detects uncovered gaps
and iteratively enhances node placement, leading to an improved coverage ratio and minimized node overlap. The
results of the hybrid meta-heuristic algorithm show improved performance compared to using PSO and ALO
separately. This approach leads to an enhanced network lifetime and energy consumption of the WSN

Accurate and efficient object detectors on resource-constrained edge devices are essential for realtime
space applications such as satellite monitoring and deep sky exploration. However, Deep Learning (DL)
models face difficulties when implemented in environments with limited memory and computational capabilities.
To address this problem, this study proposes two lightweight architectures based on YOLOv10s and YOLOv10n
that integrate ShuffleNetv2 and MobileNetv2. The Shuffle architecture improves efficiency by taking the advantages
of the depthwise separable convolution and channel shuffling. The Mobile-Shuffle architecture combines Shuffle
blocks with MobileNetv2-inspired blocks to enhance feature extraction and representation while maintaining low
computational cost. This work evaluates both architectures on the augmented DeepSapceYOLODataset dataset,
which contains a collection of annotated deep sky objects with different sizes and complexities. The evaluations
demonstrate that the Shuffle architecture achieves the highest mAP50 on both sets, the fastest inference speed,
and the smallest model size, especially in YOLOv10s architecture. While the Mobile-Shuffle architecture shows
its strength in the YOLOv10n variant by improving the detection accuracy, it comes at the cost of the increased
inference latency. These findings demonstrate that the proposed models offer a practical trade-off between accuracy,
speed, and size, making them well-suited for edge deployment in space-related scenarios.

The Internet of Things (IoT) constitutes an expanding network of interconnected gadgets that enable
intelligent systems to gather, analyze, and disseminate data. However, this rapid growth raises cyber-attack risks
due to poor configurations and outdated systems. Malware, which exploits system vulnerabilities, represents a
significant threat to the information security of IoT systems. Thus, malware detection in IoT systems is a critical
concern. Therefore, this research paper presents an IoT malware detection method based on an image dataset
and the Chi-square method as well as applying the Convolutional Neural Network (CNN) deep learning model
to detect the IoT malware. This study attempts to investigate the impact of the chi-square Feature Selection (FS)
method on the effectiveness of CNNs for identifying IoT malware, by directly applying feature selection to the
images to discern the most informative ones from the dataset before passing them to the CNN deep learning
model, demonstrating robust outcomes and validating the efficacy and robustness of the suggested approach for
identifying IoT malware. An experimental comparison was carried out between the suggested method that Involved
training the CNN on the feature-selected dataset (FS+CNN Model) and the (CNN Model) that was trained on
the full dataset and was also evaluated using the presented state-of-the-art to add to the method’s reliability.
The accuracy of the (Fs + CNN Model) reached 98.19% while its precision, recall, and F1-score were 99.52%,
95.90 %, and 97.68 %, respectively, outperforming the CNN Model’s accuracy with 94.75 %, precision with 93.00
%, recall with 91.43 % and f1-score with 90.43 %. It also outperformed the state-of-the-art evaluation with an
accuracy value of 97.93 %, a precision value of 98.64 %, a recall value of 88.73 %, and an f1-score value of 93.94 %.