Akash Sanghi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Akash Sanghi has a total of 62 co-authors. Here are the top co-authors by number of shared publications:
Gaurav Agarwal
Invertis University Bareilly,Department of CSE,India
12 shared publicationsAmit Kumar Gupta
KIET Group of Institutions,Department of Computer Applications,Ghaziabad,India
6 shared publications- Gaurav Agarwal
Invertis University
4 shared publications Atul Kumar
Invertis University, India
2 shared publications- Kamal Upreti
Department of Computer Science, CHRIST (Deemed to be University), Delhi NCR, Ghaziabad, Uttar Pradesh, India
2 shared publications Rupanshi Agarwal
Invertis University,Department of Computer Science and Engineering,Bareilly,India
2 shared publications
Latest Publications
- 5 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Hybrid CNN Architectures for Detection of Respiratory and Cardiac Diseases from Chest X-Ray Image Classification using Modern Approaches
In the present era, the advancement in the field of medicine is rapidly transforming healthcare in the diagnosis and detection of diseases at the early stages. Respiratory diseases and cardiac diseases are the two major categories of diseases that are very common these days. As the number of cases is increasing the physician, scientists who specialize in artificial intelligence have to develop some deep learning model that will help in detection of the disease using the x ray of the chest of the infected persons. In this paper we had proposed some Convolutional Neural Network model to identify the pneumonia (respiratory cases) as well as cardiac arrest (cardiovascular cases). The model architecture is based on some traditional Convolutional Neural Network models such as VGG, Alex Net, Inception, ResNet as well as some modern Convolutional Neural Network model such as Incetion V3, ResNet 50, VGG 16, VGG 19, DenseNet 201, MobileNet, Xception. In this model we had taken a combination of nearly 3300 chest X-Ray images out of which 1000 are normal and 600 are of infected chest X-Ray images. Also, a comparative study has been performed on these approaches on the basis of some of the factors such as data size, epochs, optimizer used, random state, classifier used, and best performance is obtained by DenseNet201 and its accuracy is 99.9% with 25 epochs.This study will help the researchers to develop some more effective Convolutional Neural Network model to detect the cardiovascular and respiratory diseases.
- 5 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Hybrid CNN Architectures for Detection of Respiratory and Cardiac Diseases from Chest X-Ray Image Classification using Modern Approaches
In the present era, the advancement in the field of medicine is rapidly transforming healthcare in the diagnosis and detection of diseases at the early stages. Respiratory diseases and cardiac diseases are the two major categories of diseases that are very common these days. As the number of cases is increasing the physician, scientists who specialize in artificial intelligence have to develop some deep learning model that will help in detection of the disease using the x ray of the chest of the infected persons. In this paper we had proposed some Convolutional Neural Network model to identify the pneumonia (respiratory cases) as well as cardiac arrest (cardiovascular cases). The model architecture is based on some traditional Convolutional Neural Network models such as VGG, Alex Net, Inception, ResNet as well as some modern Convolutional Neural Network model such as Incetion V3, ResNet 50, VGG 16, VGG 19, DenseNet 201, MobileNet, Xception. In this model we had taken a combination of nearly 3300 chest X-Ray images out of which 1000 are normal and 600 are of infected chest X-Ray images. Also, a comparative study has been performed on these approaches on the basis of some of the factors such as data size, epochs, optimizer used, random state, classifier used, and best performance is obtained by DenseNet201 and its accuracy is 99.9% with 25 epochs.This study will help the researchers to develop some more effective Convolutional Neural Network model to detect the cardiovascular and respiratory diseases.
- 5 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
An Intelligent Cognitive Framework Integrating SOAR Architecture with Large Language Models and AI Agents for Cyber Defence
The increasing sophistication of cyber adversaries calls for a shift in security operations from a reactive to a proactive approach in enterprise security architectures. A self-sufficient cyber defence concept based on an AI-powered SOAR (Security Orchestration, Automation, and Response) system is presented here. The proposed system features intelligent agents and machine-learning-driven threat-assessment to enable alarm triage, correlation, and incident response in real time. By injecting Large Language Models (LLMs) for contextualized reasoning and decision support, the system enhances detection accuracy, reduces human intervention, and accelerates response efficiency. The AI-driven SOAR system experimental results show that the framework significantly improves tactical agility as well as dynamic defence against evolving cyberthreats. Moreover, the model’s ability to perform adaptive learning allows it to keep learning from historical incidents, thus, it can continuously resist novel threats. The AI-SOAR solution speeds up incident response and recovery times while lowering false positives by implementing context-aware reasoning. In comparison to traditional rule-based SOAR systems, study results employing baseline intrusion detection datasets like CICIDS-2017 show enhanced detection accuracy, efficiency in operation, and scalability. Thus, the suggested AI-SOAR framework offers a self-learning, intelligent, and transparent approach to contemporary cyber threat management, filling the gap across automated and autonomous in cyberspace.
- 5 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
An Intelligent Cognitive Framework Integrating SOAR Architecture with Large Language Models and AI Agents for Cyber Defence
The increasing sophistication of cyber adversaries calls for a shift in security operations from a reactive to a proactive approach in enterprise security architectures. A self-sufficient cyber defence concept based on an AI-powered SOAR (Security Orchestration, Automation, and Response) system is presented here. The proposed system features intelligent agents and machine-learning-driven threat-assessment to enable alarm triage, correlation, and incident response in real time. By injecting Large Language Models (LLMs) for contextualized reasoning and decision support, the system enhances detection accuracy, reduces human intervention, and accelerates response efficiency. The AI-driven SOAR system experimental results show that the framework significantly improves tactical agility as well as dynamic defence against evolving cyberthreats. Moreover, the model’s ability to perform adaptive learning allows it to keep learning from historical incidents, thus, it can continuously resist novel threats. The AI-SOAR solution speeds up incident response and recovery times while lowering false positives by implementing context-aware reasoning. In comparison to traditional rule-based SOAR systems, study results employing baseline intrusion detection datasets like CICIDS-2017 show enhanced detection accuracy, efficiency in operation, and scalability. Thus, the suggested AI-SOAR framework offers a self-learning, intelligent, and transparent approach to contemporary cyber threat management, filling the gap across automated and autonomous in cyberspace.
- 5 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Advanced Energy-Efficient Clustering Protocols for IoT-based Wireless Sensor Networks: A Review and Future Direction
Wireless sensor networks (WSNs) form a basic technology of a wide range of Internet of Things (IoT) applications such as precision agriculture, environmental monitoring and industrial automation. However, due to the limited energy resources of individual sensor nodes, the scalability and last-mile deployment are great challenges. Consequently, clustering-based routing has become a feasible methodology to minimize the energy consumption of these networks. The present study brings Swarm Intelligence, Machine Learning and Predictive Threshold Models in the foreground and conducts a complete review of different clustering methods targeting energy efficiency in WSNs enabled with IoT. Each method is discussed in terms of the theory, implementation requirements, and the empirical performance metrics. Moreover, in this research, a hybrid and three layer architecture involving the combination of zone based machine learning clustering, the swarm intelligence based cluster head selection and the adaptive predictive modeling for the data transmission is proposed. Comparative evaluations in recent literature show that this joint strategy significantly extends network lifetime, reduces the communication overhead and improves the data throughput. Finally, the study provides potential future research directions to design intelligent, scalable and resilient WSNs that can adapt to the changing needs of the applications, while continuing to operate efficiently.
- 5 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Advanced Energy-Efficient Clustering Protocols for IoT-based Wireless Sensor Networks: A Review and Future Direction
Wireless sensor networks (WSNs) form a basic technology of a wide range of Internet of Things (IoT) applications such as precision agriculture, environmental monitoring and industrial automation. However, due to the limited energy resources of individual sensor nodes, the scalability and last-mile deployment are great challenges. Consequently, clustering-based routing has become a feasible methodology to minimize the energy consumption of these networks. The present study brings Swarm Intelligence, Machine Learning and Predictive Threshold Models in the foreground and conducts a complete review of different clustering methods targeting energy efficiency in WSNs enabled with IoT. Each method is discussed in terms of the theory, implementation requirements, and the empirical performance metrics. Moreover, in this research, a hybrid and three layer architecture involving the combination of zone based machine learning clustering, the swarm intelligence based cluster head selection and the adaptive predictive modeling for the data transmission is proposed. Comparative evaluations in recent literature show that this joint strategy significantly extends network lifetime, reduces the communication overhead and improves the data throughput. Finally, the study provides potential future research directions to design intelligent, scalable and resilient WSNs that can adapt to the changing needs of the applications, while continuing to operate efficiently.
- 14 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
E-TAN: Emoji Enhanced Two-Level Attention Network
Emojis have become an essential component of online communication in the digital age, acting as non-verbal indicators that transmit tone, meaning, and emotions more effectively than traditional text. Although emojis are extensively used, incorporating them into natural language processing (NLP) poses unique challenges due to their symbolic nature and context-dependent semantics. This study investigates the use of emojis in sentiment and emotion analysis using BERT (Bidirectional Encoder Representations from Transformers), a state-of-the-art language representation model based on the Transformer architecture. By leveraging BERT's contextual understanding capabilities, this work aims to bridge the semantic gap between text and emojis and enhance machine comprehension of emoji-enriched content. The model has been trained and evaluated on datasets that comprise a range of emojis in both short and long texts in order to identify and classify emotional patterns. The comprehensive study demonstrates how BERT can effectively learn and infer the emotional context of emojis inside textual data, hence improving sentiment classification accuracy. This work promotes more sophisticated and human-like comprehension in AI systems and advances the developing domains of affective computing and text analytics by demonstrating how deep learning models may be adapted to interpret modern digital language.
- 14 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
E-TAN: Emoji Enhanced Two-Level Attention Network
Emojis have become an essential component of online communication in the digital age, acting as non-verbal indicators that transmit tone, meaning, and emotions more effectively than traditional text. Although emojis are extensively used, incorporating them into natural language processing (NLP) poses unique challenges due to their symbolic nature and context-dependent semantics. This study investigates the use of emojis in sentiment and emotion analysis using BERT (Bidirectional Encoder Representations from Transformers), a state-of-the-art language representation model based on the Transformer architecture. By leveraging BERT's contextual understanding capabilities, this work aims to bridge the semantic gap between text and emojis and enhance machine comprehension of emoji-enriched content. The model has been trained and evaluated on datasets that comprise a range of emojis in both short and long texts in order to identify and classify emotional patterns. The comprehensive study demonstrates how BERT can effectively learn and infer the emotional context of emojis inside textual data, hence improving sentiment classification accuracy. This work promotes more sophisticated and human-like comprehension in AI systems and advances the developing domains of affective computing and text analytics by demonstrating how deep learning models may be adapted to interpret modern digital language.
- 7 November 2025
- Published by Taylor & Francis
- p. 1-18
Cognitive robotics for social, educational, and therapeutic applications
Cognitive robotics is an interdisciplinary field that combines artificial intelligence, robotics, and cognitive science to create intelligent robotic systems with perception, reasoning, and decision-making capability. This chapter explores the role of cognitive robotics in social, educational, and therapeutic applications. It also examines the potential benefits, challenges, and future prospects. The study reviews existing literature and analyzes current technological advancements, and interprets findings from experimental and real-world implementations. Cognitive robots improve human interactions, learning outcomes, and therapy effectiveness, but challenges like cost and ethics persist. The discussion covers the ethical considerations, limitations, and future recommendations. The evidence leans toward benefits like reduced loneliness and personalized learning, and includes privacy concerns and potential human replacement controversies. The results suggest that the cognitive robots significantly enhance human interactions, learning processes, and therapeutic interventions, and offer innovative solutions for various domains.
- 7 November 2025
- Published by Taylor & Francis
- p. 1-18
Cognitive robotics for social, educational, and therapeutic applications
Cognitive robotics is an interdisciplinary field that combines artificial intelligence, robotics, and cognitive science to create intelligent robotic systems with perception, reasoning, and decision-making capability. This chapter explores the role of cognitive robotics in social, educational, and therapeutic applications. It also examines the potential benefits, challenges, and future prospects. The study reviews existing literature and analyzes current technological advancements, and interprets findings from experimental and real-world implementations. Cognitive robots improve human interactions, learning outcomes, and therapy effectiveness, but challenges like cost and ethics persist. The discussion covers the ethical considerations, limitations, and future recommendations. The evidence leans toward benefits like reduced loneliness and personalized learning, and includes privacy concerns and potential human replacement controversies. The results suggest that the cognitive robots significantly enhance human interactions, learning processes, and therapeutic interventions, and offer innovative solutions for various domains.
- Open Access
- 29 October 2025
- Published by SPS Educational Trust in International Journal of Research and Review in Applied Science, Humanities, and Technology
Federated Learning in the Era of Decentralized Intelligence: Challenges and Opportunities
Federated Learning (FL) is a new paradigm in distributed machine learning that allows numerous participants to train a common model without exchanging raw data. This strategy improves data privacy and security, addressing major concerns in sensitive domains including healthcare, banking, and mobile applications. Unlike traditional centralized learning, FL moves model training closer to the data source, reducing data leakage risks and boosting regulatory compliance (e.g., GDPR, HIPAA). This paper provides a complete overview of federated learning, including its core architecture, distinct types (horizontal, vertical, and transfer learning), enabling technologies (such as secure multiparty computation, differential privacy, and homomorphic encryption), and real-world applications. The paper also examines the fundamental issues encountered by FL, such as data heterogeneity, communication overhead, and vulnerability to adversarial assaults. The report concludes with interesting research prospects for increasing model performance, security, and scalability in federated environments.
- 29 October 2025
- Published by SPS Educational Trust in International Journal of Research and Review in Applied Science, Humanities, and Technology
Open AccessFederated Learning in the Era of Decentralized Intelligence: Challenges and Opportunities
Federated Learning (FL) is a new paradigm in distributed machine learning that allows numerous participants to train a common model without exchanging raw data. This strategy improves data privacy and security, addressing major concerns in sensitive domains including healthcare, banking, and mobile applications. Unlike traditional centralized learning, FL moves model training closer to the data source, reducing data leakage risks and boosting regulatory compliance (e.g., GDPR, HIPAA). This paper provides a complete overview of federated learning, including its core architecture, distinct types (horizontal, vertical, and transfer learning), enabling technologies (such as secure multiparty computation, differential privacy, and homomorphic encryption), and real-world applications. The paper also examines the fundamental issues encountered by FL, such as data heterogeneity, communication overhead, and vulnerability to adversarial assaults. The report concludes with interesting research prospects for increasing model performance, security, and scalability in federated environments.
- 25 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Integrating Emotion Recognition, Vehicle Detection & Predictive Safety: A Comprehensive Survey for Intelligent Transportation Systems
Transportation safety constitutes a paramount concern in modern society. Although advanced driver assistance systems (ADAS) provide substantial assistance, their efficacy is constrained by rule-based reasoning and a lack of contextual awareness. This survey examines contemporary advancements in artificial intelligence, emphasizing emotion recognition, vehicle detection and predictive safety to uncover potential avenues for the integration of these fields into intelligent, real-time safety systems. The manuscript evaluates multimodal methodologies, elucidates deficiencies within current systems, and suggests a cohesive architectural framework capable of improving safety outcomes under variable driving conditions.
- 25 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Integrating Emotion Recognition, Vehicle Detection & Predictive Safety: A Comprehensive Survey for Intelligent Transportation Systems
Transportation safety constitutes a paramount concern in modern society. Although advanced driver assistance systems (ADAS) provide substantial assistance, their efficacy is constrained by rule-based reasoning and a lack of contextual awareness. This survey examines contemporary advancements in artificial intelligence, emphasizing emotion recognition, vehicle detection and predictive safety to uncover potential avenues for the integration of these fields into intelligent, real-time safety systems. The manuscript evaluates multimodal methodologies, elucidates deficiencies within current systems, and suggests a cohesive architectural framework capable of improving safety outcomes under variable driving conditions.
- 16 July 2025
- Published by Springer Nature
- p. 13-23
Exploring Techniques for Detecting Copy-Move Forgeries in Digital Images: A Concise Survey
- 16 July 2025
- Published by Springer Nature
- p. 13-23
Exploring Techniques for Detecting Copy-Move Forgeries in Digital Images: A Concise Survey