Computer Science Review最新文献

{"title":"A survey on AI-empowered task-oriented sensing, communication, and computation in 6G networks","authors":"Yuxin Zhang,&nbsp;Xingwei Wang,&nbsp;Xuewen Luo,&nbsp;Xinyue Pei,&nbsp;Fuliang Li,&nbsp;Donghong Han,&nbsp;Tianyu Li","doi":"10.1016/j.cosrev.2026.100899","DOIUrl":"10.1016/j.cosrev.2026.100899","url":null,"abstract":"<div><div>Deeply empowered by artificial intelligence (AI) technologies, integrated sensing, communication, and computation (ISCC) architectures have emerged as a critical component in the sixth-generation (6G) mobile communication networks. This survey is conducted to provide a thorough analysis of AI-empowered task-oriented sensing, communication, and computation (AI-TSCC) in 6G networks. At first, the intrinsic relationships among sensing, communication, and computation are systematically reviewed, and the limitations of current technologies are identified. Then, the AI-empowered methods for performance improvement in ISCC are analyzed, including AI-based TSCC and AI-assisted TSCC. Next, the tasks are classified into three categories according to their characteristics. Focusing on the optimization performance of AI-TSCC systems, model-driven metrics, task execution metrics, and other relevant metrics are introduced. Furthermore, existing technical bottlenecks and future research directions are summarized to provide theoretical and practical guidelines for building efficient 6G networks.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100899"},"PeriodicalIF":12.7,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146014531","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Isogeny-based cryptography: A comprehensive review on advancements, analysis of attacks, and future directions","authors":"Dheerendra Mishra,&nbsp;Rohit Raj Sharma","doi":"10.1016/j.cosrev.2025.100865","DOIUrl":"10.1016/j.cosrev.2025.100865","url":null,"abstract":"<div><div>Recent developments in quantum computing present significant challenges to traditional cryptographic approaches, especially those relying on integer factorization and discrete logarithms. To counter these challenges, a variety of post-quantum cryptographic (PQC) methods have been proposed, including cryptographic systems based on lattices, hashing, codes, multivariate equations, and isogenies. Each of these is based on complex mathematical problems that are challenging to solve even with quantum computers. Among them, isogeny-based cryptography has garnered considerable attention for its small key sizes, making it a leading candidate for developing efficient post-quantum cryptographic primitives. This article provides an in-depth review of isogeny-based cryptographic techniques, emphasizing key exchange, zero-knowledge proofs (ZKP), and signature protocols. It begins by elucidating the foundational concepts of elliptic curves, isogenies, and the primary protocols utilizing them. The discussion then shifts to the challenges that current schemes face and examines potential research directions aimed at improving security, scalability, and functionality.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100865"},"PeriodicalIF":12.7,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145689752","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Particle swarm optimization: A survey of innovations over the last 10 years","authors":"Angel Casas-Ordaz ,&nbsp;Eduardo H. Haro ,&nbsp;Luis A. Beltran ,&nbsp;Omar Alvarez ,&nbsp;Seyed Jalaleddin Mousavirad ,&nbsp;Marco Pérez-Cisneros ,&nbsp;Diego Oliva","doi":"10.1016/j.cosrev.2026.100910","DOIUrl":"10.1016/j.cosrev.2026.100910","url":null,"abstract":"<div><div>The Particle Swarm Optimization (PSO) is a well-known metaheuristic algorithm whose impact on optimization is invaluable. Its simple design, proposed in 1995, has been the inspiration for several related approaches, some of which are pillars in the literature. Additionally, the PSO has parameters that can be tuned from different perspectives, which has led to its application in numerous areas and to its hybridization with most of the evolutionary methods in the state-of-the-art. These features have highlighted the need to present ongoing review work that offers an overview of PSO’s contributions to the identified research areas. Nonetheless, the algorithm’s relevance has made it increasingly difficult to summarize its overall impact in a single article. Furthermore, many related works that limit their scope tend to categorize the PSO literature into mixed categories, which may confuse readers rather than guide them toward a deeper understanding of the approach. Therefore, this article presents an exhaustive survey of the PSO, highlighting three remarkable aspects. First, the taxonomy is limited to the last decade to illustrate the scheme’s impact in the contemporary world. Secondly, a structure of seven main categories is used to define the areas where the optimizer has contributed. Finally, different statistical analyses enable the reader to understand the evolution of the method and its future directions, which is particularly important for new PSO research. One of the principal purposes of this survey is to serve as an expanded summary of the latest advances in the PSO algorithm over a 10-year period. This study is expected to generate interest among researchers in the field, providing a comprehensive overview of the PSO algorithm’s performance and serving as a valuable resource for experienced researchers.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100910"},"PeriodicalIF":12.7,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146134549","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AIoT for human emotion recognition: Potentials, challenges, and healthcare applications","authors":"Shumayla Yaqoob ,&nbsp;Farman Ullah ,&nbsp;Najah AbuAli ,&nbsp;Hafeez Anwar ,&nbsp;Nasir Saeed ,&nbsp;Mohammad Hayajneh","doi":"10.1016/j.cosrev.2025.100859","DOIUrl":"10.1016/j.cosrev.2025.100859","url":null,"abstract":"<div><div>Emotions are critical human behavior and cognition drivers, influencing communication, decision-making, and well-being. Emotion recognition (ER) is the computational identification of emotional states, and it has therefore gained considerable attention across various fields, including human-computer interaction, mental health, and intelligent systems. This review article synthesizes recent advancements in ER enabled by the integration of the Internet of Things (IoT) and Artificial Intelligence (AI), collectively termed AIoT, with a specific focus on healthcare applications. We highlight IoT-based sensing technologies, including wearables, ambient sensors, and mobile devices, which enable continuous and non-intrusive monitoring of emotions through multimodal signals such as facial expressions, speech, EEG, ECG, and GSR. A comprehensive taxonomy is proposed that organizes sensing modalities, datasets, pre-processing methods, learning algorithms, and application domains. Both traditional machine learning methods (e.g., SVM, Random Forests) and modern deep learning approaches (e.g., CNNs, LSTMs, Transformers) are evaluated for their ability to effectively handle complex emotional data. The integration of AI and IoT is presented as essential for developing scalable, real-time, and context-sensitive emotion-aware systems for healthcare applications. We discuss key challenges such as data heterogeneity, privacy, interpretability, and limited labeled datasets along with future directions such as edge computing, federated learning, and explainable AI. This synthesis aims to guide the development of robust, personalized, AIoT-enabled emotion-aware healthcare systems.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100859"},"PeriodicalIF":12.7,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145613574","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Transformers meet CNNs: A comprehensive review and benchmarking of deep learning architectures for brain tumor classification in MRI","authors":"Noor ul Ain ,&nbsp;Sajid Ali Khan ,&nbsp;Suliman Aladhadh ,&nbsp;Usama Mir ,&nbsp;Muhammad Ramzan","doi":"10.1016/j.cosrev.2026.100897","DOIUrl":"10.1016/j.cosrev.2026.100897","url":null,"abstract":"<div><div>Early and accurate detection of brain tumors in MRI images is important for increasing patient survival rates. The latest trends in Deep Learning (DL) have revolutionized medical imaging analysis. This paper presents a systematic review of state-of-the-art methodologies, including Convolutional Neural Networks (CNNs), Transformers, Graph Neural Networks (GNNs), and Explainable AI (XAI). It also summarizes the information on pre-processing techniques, commonly available datasets, and evaluation metrics. Following that, experimental validation is performed on various DL models, including CNNs, a Custom Deep CNN (for Ablation), transfer learning models (VGG16, ResNet50, EfficientNetB0), and a Swin Transformer. The transformer achieved the superior Mean ± Std accuracy (0.97 ± 0.02), Precision 99%, Recall 99% and F1-score (98%) across 5 Runs. Moreover, the Attention Maps of the Swin Transformer are evaluated, providing insight into the decision-making process of DL models. Finally, the study also highlights existing challenges and outlines future research directions, including federated learning, self-supervised approaches, and lightweight hybrid architectures to build scalable, interpretable diagnostic models.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100897"},"PeriodicalIF":12.7,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146033308","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Potential of artificial intelligence in deepfake media: From generation to detection mechanisms, state-of-the-art, and challenges","authors":"Shubham Sharma,&nbsp;Arvind Selwal","doi":"10.1016/j.cosrev.2025.100866","DOIUrl":"10.1016/j.cosrev.2025.100866","url":null,"abstract":"<div><div>Artificial intelligence (AI) plays an important role in the generation of deepfakes by leveraging advanced machine learning models to create hyper-realistic synthetic media across visual, audio, and multimodal formats. The rapid evolution of deepfake technologies, alongside the exponential growth of digital media, demands a comprehensive and critical examination of current capabilities and challenges. Although the concept of media manipulation is not new, the sophistication and accessibility of AI-driven deepfakes present significant threats of misinformation to society and hence cause societal manipulation. This manuscript presents a systematic review of deepfake generation and detection techniques from 2017 to 2025, highlighting the progression of generative models and evaluating detection strategies. The main focus of this work is on the state-of-the-art (SOTA) techniques using adversarial networks, vision transformers (ViTs), attention mechanisms, hybrid learning frameworks, and ensemble models. The study thoroughly examines the benefits and drawbacks of existing methods. It also points out how vulnerable detection systems are to adversarial attacks and compares modern methods with traditional forensic and heuristic approaches. The paper critically analyzes the strengths and limitations of existing models, underscores the susceptibility of detection systems to adversarial attacks, and contrasts contemporary approaches with traditional forensic and heuristic-based methods. In addition to technical insights, the review puts a major focus on practical concerns such as scalability, regulatory frameworks, and the broader societal impact of the deepfake technology. By including benchmark datasets, standard tools, performance evaluation metrics, and relevant policy discussions, the manuscript presents a forward-looking perspective on the ongoing arms race between deepfake generation and detection. The study ends by highlighting the need for strong, flexible, and understandable detection systems, backed by effective policy measures, to reduce the growing risks posed by deepfakes.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100866"},"PeriodicalIF":12.7,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145689751","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Artificial intelligence and machine learning techniques in solid waste management: A sustainable way toward future","authors":"Soghra Nashath Omer ,&nbsp;Panchamoorthy Saravanan ,&nbsp;Pramilaa Kumar ,&nbsp;M. Moniga ,&nbsp;R. Rajeshkannan ,&nbsp;Madhavi Reddy ,&nbsp;M. Rajasimman ,&nbsp;S. Venkat Kumar","doi":"10.1016/j.cosrev.2025.100889","DOIUrl":"10.1016/j.cosrev.2025.100889","url":null,"abstract":"<div><div>Solid Waste Management (SWM) constitutes a significant challenge confronting both developed and developing countries. A crucial element of effective solid waste management is ensuring that waste bins is public spaces are adequately filled prior to the commencement of the subsequent cleaning cycle. Failure to do so can result in various hazards, including unsightly litter and unpleasant odours, which may contribute to the proliferation of diseases. Furthermore, the rapid growth of the population has markedly strained the existing SWM infrastructure, particularly in terms of sanitation facilities. The indiscriminate disposal of garbage in public area leads to environmental pollution. To mitigates waste-related issues and uphold public health standards, the implementation of a comprehensive SWM system is essential. It is important to recognize that the necessity for effective waste management extends beyond merely the collection and disposal of waste. And also, the study examines the implementation of Artificial Intelligence (AI) and Machine Learning (ML) applications in SWM, evaluate the performance of these AI and ML applications investigates the associated benefits and challenges and offers recommendations for best practices aimed at optimizing resource efficiency to enhance economic, environmental and social outcomes. The research will be advantageous for scholars, government entities, policy-makers, and various organizations involved in waste management, as it seems to evaluate current recycling rates, minimize reliance on manual labour decrease operational costs, enhance efficiency and fundamentally transform the methodologies employed in the solid waste management.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100889"},"PeriodicalIF":12.7,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145884462","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Low-rank completion for motion capture data recovery: Approaches, constraints, and algorithms","authors":"Souad Mohaoui ,&nbsp;Andrii Dmytryshyn","doi":"10.1016/j.cosrev.2025.100878","DOIUrl":"10.1016/j.cosrev.2025.100878","url":null,"abstract":"<div><div>Motion capture (MoCap) systems are indispensable tools across fields such as biomechanics, computer animation, human-robot interaction, and clinical gait analysis, owing to their ability to accurately record and analyze human movement in 3D space. Marker-based systems use reflective markers attached to subjects and video recordings to track human movement. The tracking requires markers to be detected in the video, which is not always possible due to occlusions, sensor failures, and limited camera coverage. These issues create gaps in recorded trajectories, compromising data integrity and making the motion difficult to utilize in practical applications. Therefore, a wide range of MoCap data completion techniques has been proposed to reconstruct missing trajectories while preserving the realism and dynamics of human movement. Human motion data exhibits a low-rank property due to the inherent repetitive nature of human movement as well as the correlations between joints and markers, enforced by the skeletal structure and biomechanical constraints. Low-rank completion techniques exploit this property to reconstruct missing marker positions. This paper reviews state-of-the-art low-rank completion methods for MoCap data completion, focusing specifically on optimization-based low-rank methods. These optimization approaches directly address the missing data completion problem through optimization formulations. We examine two main aspects: kinematic priors, which embed anatomical constraints, joint dependencies, and motion smoothness, and low-rank priors, which exploit inter-marker correlations through matrix and tensor formulations. We further evaluate optimization algorithms for solving these completion problems, such as alternating minimization, proximal algorithms, ADMM, and hybrid schemes, as well as the datasets and tools commonly used in the literature.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100878"},"PeriodicalIF":12.7,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145938393","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A survey of transformer networks for time series forecasting","authors":"Jingyuan Zhao ,&nbsp;Fulin Chu ,&nbsp;Lili Xie ,&nbsp;Yunhong Che ,&nbsp;Yuyan Wu ,&nbsp;Andrew F. Burke","doi":"10.1016/j.cosrev.2025.100883","DOIUrl":"10.1016/j.cosrev.2025.100883","url":null,"abstract":"<div><div>Time-series data support critical functions in forecasting, anomaly detection, resource optimization, and real-time decision making across finance, healthcare, energy systems, and networked computing. Classical statistical approaches and early deep-learning architectures (RNNs, LSTMs, CNNs) have achieved notable progress, yet they exhibit structural limitations: recurrent models struggle with long-range dependencies, convolutional models require deep stacking to enlarge receptive fields, and both scale suboptimally to high-dimensional and high-volume data. Transformer architectures—characterized by global attention mechanisms, flexible temporal receptive fields, and growing adoption within foundation-model paradigms—have consequently gained increasing prominence for modern time-series analysis. Drawing on a systematic review of IEEE Xplore, ACM Digital Library, and Scopus (2020–2025), this survey offers a unified, theoretically grounded synthesis of Transformer-based methods for time-series learning. The survey summarizes defining characteristics of time-series data and analyzes core architectural elements, including attention formulations, encoder–decoder structures, hyperparameter design, and domain-specific adaptations. An architecture-centered and task-aware taxonomy is presented to organize recent advances across forecasting, representation learning, anomaly detection, and multimodal fusion. Persistent challenges—spanning computational scalability, data-efficiency constraints, distributional heterogeneity, overfitting risks, hyperparameter instability, interpretability, and reproducibility—are examined in depth. A forward-looking research agenda is outlined, highlighting opportunities in physics-informed architectural design, hybrid neural–mechanistic modeling, resource-efficient real-time inference, multi-resolution spatiotemporal learning, and emerging human–AI collaborative paradigms. By consolidating these methodological developments, this survey aims to provide a structured reference point for ongoing research on Transformer models for time-series machine learning.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100883"},"PeriodicalIF":12.7,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145785081","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AI-driven semantic information retrieval for Arabic language- systematic literature review","authors":"Amany A. Slamaa","doi":"10.1016/j.cosrev.2026.100920","DOIUrl":"10.1016/j.cosrev.2026.100920","url":null,"abstract":"<div><div>Despite the rapid growth of AI-powered text, Arabic's content remains semantically underrepresented and don’t have enough interpretation for machines. Limited synthesizing of how semantic resources relate to AI-driven retrieval or generative systems. Limited review studies introduced Arabic semantic web since the last decade, without projection of evolution. Besides, the lack of practical implications and synthesis for future research directions in AI environments makes a gap for who cares about Arabic semantic information retrieval in artificial intelligence environment. Thereby, the study demonstrates the contributions, challenges, and milestones in structuring Arabic semantic information retrieval (ASIR), and explores how these advancements have influenced the development of Arabic language models and AI contributions like LLMs and Chatbots over the past two decades. Integrating Systematical Literature Review, PRISMA, Design Science Research and AMSTAR-2 provide answers to multiple research questions tracking advancements from 2000 to 2025. As a result, ASIR has developed from rule-based to neural and LLM-driven systems, attaining over a 35 % rise in accuracy and ten dialects merged. Key progresses incorporate Arabic DBpedia integration, ontology frameworks, and models like AraBERT for enhanced contextual information retrieval. Additionally, dialectal NLP has been developed for applications such as Chatbots and LLMs. Although challenges in standardized benchmarks, dataset diversity, ASIR's assistance in intelligent Chatbots, adaptive e-learning, etc., this study serves as a valuable resource for researchers, developers, and policymakers aiming to advance linguistically inclusive and semantically rich AI solutions for Arabic-speaking communities. This study is regarded as a roadmap, diagnostic tool, and strategic guide for enhancing Arabic AI systems with semantic depth, linguistic inclusivity, and real-world impact.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100920"},"PeriodicalIF":12.7,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146134551","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}