University of Isfahan & Iranian Society of CryptologyJournal of Computing and Security2322-446013120260101Vol.13, No. 1, 202619730696ENJournal Article20260817https://jcomsec.ui.ac.ir/article_30696_394c77dfe29aefbde01f1309dc9ec27d.pdfUniversity of Isfahan & Iranian Society of CryptologyJournal of Computing and Security2322-446013120260401An Ant-Colony-Optimization-Based Method for Community-Aware Link Prediction in Social Networks1213029110.22108/jcs.2026.145007.1166ENSeyed MohammadHosseiniFaculty of Computer Engineering, University of Isfahan, Iran.AfsanehFatemiFaculty of Computer Engineering, University of Isfahan, Iran.MaryamHosseini-PozvehDepartment of Computer Engineering, Shahreza Campus, University of Isfahan, Iran.Journal Article20250421Link prediction is one of the primary issues in the social network analysis domain, focusing on predicting the emergence of future links. Among state-of-the-art approaches, metaheuristic-based methods stand out, as they can combine existing heuristic measures for link formation with greedy exploration, thereby improving the effectiveness of future link prediction. Many of these heuristics are designed based on the structural characteristics of social networks. However, only a limited number of methods have incorporated community structure, a fundamental property of social networks. In this article, a novel link prediction method is introduced by integrating the concept of community structure into the ant colony optimization algorithm. Specifically, three common neighbor-based heuristics are extended to consider the community structure and the distinct nature of intra-community and inter-community relationships. These enhanced heuristics are then utilized in the ant colony optimization search mechanism. The effectiveness of the proposed method is evaluated against several benchmark techniques using nine real-world datasets. The results demonstrate the outperformance of the presented approach considering AUC and precision comparison metrics.https://jcomsec.ui.ac.ir/article_30291_d547af18a0579cc4c09ef0334ba719a5.pdfUniversity of Isfahan & Iranian Society of CryptologyJournal of Computing and Security2322-446013120260401Integrating VGG16 with the ACIMD Protocol to Enhance Security and Reliability in ECG-Based Remote Authentication Systems23343029210.22108/jcs.2026.145712.1173ENNargesEshaghiDepartment of Mathematics, Tafresh University, Iran.MohammadHabibiDepartment of Mathematics, Tafresh University, Iran.NasourBagheriCPS2 Lab, Department of Communication, Faculty of Electrical Engineering, Shahid Rajaee Teacher Training University, Lavizan, Tehran, Iran.
Electronics Research Institute, Sharif University of Technology, Tehran, Iran.AliKhatibiDepartment of Mathematics, Tafresh University, Iran.Journal Article20250618The security of Implantable Medical Devices (IMDs) is of paramount importance, as unauthorized access can lead to life-threatening consequences. Electrocardiogram (ECG) signals present a promising biometric modality for authentication due to their inherent uniqueness and capability for liveness detection capability. However, wireless ECG-based systems are vulnerable to relay attacks, necessitating robust proximity verification. This study proposes a novel, integrated authentication framework that addresses both user identification and physical proximity. We enhance the established Access Control for Implantable Medical Devices (ACIMD) distance-bounding protocol by incorporating a fine-tuned VGG16 deep convolutional neural network to achieve high-accuracy ECG biometric verification. The system was rigorously evaluated using the MIT-BIH Arrhythmia Database. The integrated model achieved an overall authentication accuracy of 99.45%, surpassing the baseline ACIMD protocol accuracy of 97.82%. This represents an average improvement of 1.63% across various physical distance thresholds. Although integrating VGG16 increased the total authentication time from 0.0113 s to 0.0437 s, this remains well within acceptable limits for real-time medical applications. Crucially, we provide new evidence for improved system reliability, demonstrating superior robustness against signal noise compared to the baseline. The proposed system effectively balances high biometric fidelity with stringent physical-layer security, offering a comprehensive solution for secure remote authentication in critical applications such as IMDs.https://jcomsec.ui.ac.ir/article_30292_5a9234161807677efec69b27991a1640.pdfUniversity of Isfahan & Iranian Society of CryptologyJournal of Computing and Security2322-446013120260401Behavioral Loyalty Prediction from Customer Reviews in E-Commerce: A Deep Learning and Explainable AI Approach35573044110.22108/jcs.2026.147672.1188ENPedramKazemi EsfeFaculty of Computer Engineering, University of Isfahan, Isfahan, Iran.FakhroddinNoorbehbahaniFaculty of Computer Engineering, University of Isfahan, Isfahan, Iran.Journal Article20251203Understanding and predicting customer loyalty has become essential in increasingly competitive digital markets, yet conventional behavioral models, such as RFM, remain limited by their reliance on retrospective transactional data. Meanwhile, vast volumes of rich, interactive customer feedback, particularly online reviews, remain chronically underutilized for loyalty prediction. This study introduces a novel deep learning framework that predicts behavioral loyalty directly from customer review texts, eliminating the need for historical purchase records. Using a large e-commerce dataset of 58,841 samples, we combine advanced transformer-based language models (BERT, RoBERTa, ALBERT, GPT, and Qwen) with structured features and employ a genetic algorithm to optimize across 19,440 configurations of loyalty labeling schemes, neural architectures, text embeddings, and feature combinations. The resulting model achieves an accuracy of 0.92 and an F1-score of 0.90, substantially outperforming baseline machine learning methods and non-transformer text embedding approaches. Beyond performance, the integration of SHAP-based explainability provides transparent insights into the linguistic and behavioral cues driving loyalty predictions. These findings demonstrate that customer reviews contain powerful behavioral signals that can be transformed into actionable intelligence for real-time CRM, customer retention, competitive analysis, and multi-channel personalization, particularly in contexts where RFM data are incomplete or unavailable. This work offers both a methodological advancement in loyalty prediction and a practical, scalable tool for organizations seeking to leverage interactive feedback to build stronger, more proactive customer relationships.https://jcomsec.ui.ac.ir/article_30441_564d382c0f3e2a08391c5485119d97ee.pdfUniversity of Isfahan & Iranian Society of CryptologyJournal of Computing and Security2322-446013120260401Collaborative Fog Computing for Computation Reuse Based on Popularity of Services in IoT59713049410.22108/jcs.2026.147056.1182ENMarzieh SadatZahediniaFaculty of Electrical and Computer Engineering, University of Birjand, Birjand, Iran.Journal Article20251015Fog computing has emerged as a critical paradigm for processing data closer to the edge. Managing fog nodes effectively is essential for optimizing performance. This study explores a new strategy for reducing response time in IoT, focusing on computation reuse. Computation reuse is a technique used in edge computing to prevent redundant processing. This approach helps in eliminating repetitive calculations, enhancing the capacity of fog computing resources. Although computation reuse saves a significant number of computational resources when the result is cached, it also incurs extra computational costs due to operations like hashing and similarity calculation. It may be beneficial to directly process tasks without searching for the cached results, especially when the probability of successfully finding the target result is low. A novel computation reuse approach among interconnected fog nodes is proposed, leveraging parallel processing and popularity-aware caching to minimize response times and optimize storage. Our method utilizes Locality Sensitive Hashing (LSH) to detect similar computation requests and executes parallel processes for cache lookup and service computation to avoid latency penalties from cache misses. Unlike prior Computation Reuse methods that serialize lookup and execution, the proposed parallel design mitigates cache-miss latency by ensuring that unsuccessful reuse attempts do not increase response time beyond that of direct execution. Evaluations demonstrate significant improvements in response time, reuse rates, energy efficiency, and scalability compared to baseline methods. This work highlights the potential of cooperative computation reuse in fog networks and lays the groundwork for future research in dynamic network topologies.https://jcomsec.ui.ac.ir/article_30494_b6717472cb8b3e2cb1ee4a84f5b5035a.pdfUniversity of Isfahan & Iranian Society of CryptologyJournal of Computing and Security2322-446013120260401A Comparative Study of AI-Generated and Human-Written Kubernetes Manifests73823057410.22108/jcs.2026.148046.1195ENNasimTabibpourDepartment of Electrical and Computer Engineering, Isfahan University of Technology, Isfahan 84156-83111, IranMohammadDakhilalianDepartment of Electrical and Computer Engineering, Isfahan University of Technology, Isfahan 84156-83111, Iran.Journal Article20260105Kubernetes is susceptible to security and quality issues when misconfigured; meanwhile, the use of generative models such as ChatGPT to produce manifests is rising. This study compares the configuration quality of human-authored manifests (1985 open-source files) with a published 2023 corpus of AI-generated manifests (98 ChatGPT-generated files aligned with a prior MSR study). Using an aligned KubeLinter-based framework, we analyze six configuration smells and report density (occurrences per 1000 raw lines) and proportion (percentage of files affected). We also add a balanced-size robustness check that subsamples 98 human-written manifests 1000 times. The results show that missing CPU and memory specifications remain more prevalent in the AI corpus under this balanced comparison. In contrast, omissions in secure execution, such as missing runAsNonRoot and readOnlyRootFilesystem, are important in both corpora. An issue-level smell-by-kind analysis of the human corpus shows that 85.16% of target-smell occurrences concentrate in Deployment and StatefulSet resources. The findings provide a focused empirical baseline and support the use of human review, static analysis, and policy enforcement before integration or deployment.https://jcomsec.ui.ac.ir/article_30574_af1c02910447c72ad84b41066b60c572.pdfUniversity of Isfahan & Iranian Society of CryptologyJournal of Computing and Security2322-446013120260401Aspect-Based Sentiment Analysis in Persian Using a Fine-Tuned LLaMA3 Model83973065610.22108/jcs.2026.145865.1175ENNiloofarRanjbarDepartment of Computer Engineering, Jam Faculty of Engineering, Persian Gulf University, Bushehr, Iran.HamedBaghbaniDepartment of Computer Engineering, Jam Faculty of Engineering, Persian Gulf University, Bushehr, Iran.Journal Article20250712Aspect-Based Sentiment Analysis (ABSA) is less developed for Persian than for English, particularly for extracting complete aspect-category-opinion-sentiment structures rather than classifying sentiment for predefined aspects. This paper presents a generative Persian ABSA framework based on fine-tuned LLaMA3-family instruction models. Given a Persian review, the model generates a JSON array of quadruples containing aspect terms, categories, opinion spans, and sentiment labels.<br />The training data are drawn from a GPT-3.5-assisted Persian ABSA corpus whose annotations were reviewed and filtered by human annotators; we therefore treat it as GPT-assisted/human-validated rather than purely human-authored. To reduce circular evaluation, we also construct an independent human-only test set from raw Persian reviews. Three annotators manually annotated 240 reviews from scratch, and a two-of-three consensus procedure produced 555 gold aspect-level labels from 239 reviews.<br />We evaluate zero-shot LLaMA3, fine-tuned LLaMA3 models, and Dorna-LLaMA3-8B-Instruct as a Persian-oriented generative baseline under the same schema, parser, and metrics. On the human-only test set, zero-shot LLaMA3 obtains 0.1426 aspect F1 and 0.1185 aspect-sentiment F1. Fine-tuned LLaMA3-CG improves the five-seed mean scores to 0.5477 0.0229 and 0.4394, 0.0212, while Dorna-CG reaches 0.5809, 0.0123 and 0.4897, 0.0194. Adding ParsiNLU substantially improves in-domain ParsiNLU performance but reduces human-only performance, revealing a domain and annotation-style trade-off.https://jcomsec.ui.ac.ir/article_30656_6e97014380c755ab884775ab6239213e.pdf