- Volume 16
- Issue 2
- (5) Articles
V. Anusuya
,
Stency V. S.
,
G. Srividhya
,
M. K. Mohammed Faizel
,
G. Arul Kumaran
,
R. Santhosh
,
P. Sherubha
Medical image analysis plays a vital role in diagnosis of diseases and the need of the day is to arrive at a simple and efficient compression technique. This paper proposes a comparative analysis of three different wavelet based medical image compression techniques. First algorithm is based on Bi-orthogonal wavelet with Parallel coding (BiWT-PC) , second is based on Haar wavelet with block coding (HWT-BC) and third algorithm is based on stationary wavelet transform with Parallel coding (SWT-PC). In this work, 3D medical image is converted into 2D slices and preprocessed using lifting scheme. Wavelet transform is applied to this preprocessed image, which divides the image into multilevel sub-bands. Then, the suitable encoding method is applied to get the compressed image. At the receiver side, the original image is recovered back by applying inverse wavelet transform and proper decoding over the compressed image. Experimentations are carried out over MRI and CT images with four quantitative metrics such as PSNR, CR, DcT and EcT. From the experimental analysis, it is observed that SWT-PC method is quite efficient since it has high PSNR and low CR.
Read More
Doi: https://doi.org/10.54216/JCIM.160201
Vol. 16
Issue. 2
PP. 01-12, (2025)
Shivanthana S.
,
Manicka Raja M.
,
Lalitha Krishnasamy
,
Karthik R.
,
R. Venkatesan
Network security faces significant challenges due to the increasing sophistication of cyber threats and the inherent class imbalance in intrusion detection datasets. To address this issue, a hybrid Boundary Equilibrium Generative Adversarial Network (BEGFAN) and Vector Quantization Variational Autoencoder (VQVAE) framework, termed BVQVAE, is proposed for Network Intrusion Detection Systems (NIDS). The framework involves preprocessing, feature extraction, and class balancing to enhance classification accuracy. Missing values are imputed, categorical features are label-encoded, and numerical attributes are normalized to ensure a structured dataset. BEGAN generates synthetic samples to mitigate class imbalance, while VQVAE extracts essential features using an encoder with quantization and a decoder for network traffic reconstruction. The model is evaluated on NSL-KDD and UNSW-NB15 datasets, achieving 82.56% accuracy, with precision, recall, G-mean, and F1-score of 86.53%, 87.65%, 86.21%, and 87.08%, respectively.
Read More
Doi: https://doi.org/10.54216/JCIM.160202
Vol. 16
Issue. 2
PP. 13-27, (2025)
D. Gowthami
,
M. Vigenesh
In recent years, federated learning (FL) has emerged as a decentralized approach to model training, enhancing data privacy by retaining data on local edge devices. While existing privacy-preserving FL frameworks, like Secure Aggregation and Homomorphic Encryption, protect data through encrypted aggregation, they often face challenges with high communication overhead, significant computational demands, and increased energy consumption. Differential privacy approaches, though customizable via privacy budgets, may also degrade model accuracy due to added noise. Addressing these limitations, we propose PrivaNet-FL (Privacy-Optimized Network for Federated Learning), an advanced FL model that optimizes privacy techniques with minimal energy costs in edge environments. PrivaNet-FL incorporates adaptive privacy and efficiency management across edge devices, such as IoT sensors and smartphones, where data processing and real-time privacy adjustments conserve energy while maintaining data security. The framework consists of three main workflows: (1) Adaptive Privacy-Scaling-modulating privacy based on device constraints, ensuring optimal energy usage through dynamic adjustments of noise in differential privacy or encryption complexity; (2) Lightweight Encryption and Secure Aggregation-employing low-complexity encryption and secure aggregation techniques, such as random masking and distributed averaging, to minimize energy without compromising data privacy; and (3) Energy-Aware Communication-Efficient FL-leveraging model compression, energy-aware scheduling, and differential privacy with controlled noise to reduce communication and energy overhead. Results demonstrate that PrivaNet-FL achieves superior model accuracy with reduced energy and communication costs compared to traditional FL methods, making it ideal for privacy-sensitive and resource-limited edge applications.
Read More
Doi: https://doi.org/10.54216/JCIM.160203
Vol. 16
Issue. 2
PP. 28-46, (2025)
R. Udaya Nirmala Mary
,
R. Santhosh
Multi-cloud computing is emerging as a transformative solution to meet the extensive computational demands of Internet of Things (IoT) devices. In networks with multiple devices and clouds, factors such as real-time computing requirements, fluctuating wireless channel conditions, and dynamic network scales introduce significant complexity. Addressing these challenges, along with the resource constraints of IoT devices, is essential for effective multi-cloud integration. This paper proposes a hybrid decision-offloading model that integrates continuous and discrete decision-making. IoT devices must learn to make coordinated decisions regarding cloud server selection, task offloading ratios, and local computation capacity. This dual-layer decision-making process involves managing both continuous and discrete variables, along with inter-device coordination, which poses considerable challenges. To address these, we introduce a probabilistic approach that transforms discrete actions, such as selecting a cloud server, into a continuous domain. We further develop a Privacy-Aware Multi-Agent Deep Reinforcement Learning (PA-MADRL) framework that combines centralized training with distributed execution. This framework minimizes overall system costs by considering energy consumption and cloud server rental fees. Each IoT device operates as an agent, autonomously learning efficient policies while alleviating its computational burden. Experimental results demonstrate that the PA-MADRL framework effectively adapts to dynamic network conditions, learning optimal offloading policies. It significantly outperforms four state-of-the-art deep reinforcement-learning models and two heuristic methods, achieving lower system costs and improved resource efficiency.
Read More
Doi: https://doi.org/10.54216/JCIM.160204
Vol. 16
Issue. 2
PP. 47-60, (2025)
Udita Malhotra
,
Ritu
Cloud-based Enterprise Resource Planning (ERP) systems have become essential to organizational operations in today's digital environment, acting as the cornerstone for managing sensitive corporate data. ERP system integration with third-party apps, however, poses serious security risks because businesses cannot afford data breaches or illegal access that could jeopardize financial records, operational integrity, and reputation. Because ERP systems are appealing targets for cybercriminals looking to obtain sensitive company data, ensuring secure data exchange is an urgent concern. ERP integration security is still a problem, despite the numerous security frameworks and measures that have been put forth. Current methods frequently fall short of effectively addressing new threats. To guarantee the safe and smooth integration of cloud-based ERP solutions with external systems, this study presents an extensible security framework. The framework reduces the risk of data interception and unauthorized access by utilizing functional and technical security measures to produce a strong, adaptable security model. To prevent data leaks and unauthorized changes, the implementation is divided into two phases: (1) securing outbound data flow from the ERP portal to third-party systems, and (2) securing inbound data flow from third-party systems into the ERP portal, which protects against malicious intrusions and breaches of data integrity.
Read More
Doi: https://doi.org/10.54216/JCIM.160205
Vol. 16
Issue. 2
PP. 61-76, (2025)
