Abdulbaqi Khashea A. Al-Hadeethi1 and Ezzeddine Zagrouba2, 1LIMTIC, LR16ES06 Research Laboratory, Higher Institute of Computer Science el Manar, Tunis "ISI" Tunis El-Manar University , Tunis , Tunisia, 2LIMTIC, LR16ES06 Research Laboratory, Higher Institute of Computer Science el Manar, Tunis "ISI" Tunis El-Manar University , Tunis , Tunisia
Trust in authority is foundational to the successful adoption and utilization of cloud data-based information in data mining technology. Organizations entrust their sensitive data to cloud service providers, relying on them to ensure security. The technique include in this methods are Data collection and pre-processing, Feature extraction, Feature selection, Classification, Trust authentication, Encryption and decryption. To the raw data undergoes pre-processing steps include, Dimensionality reduction, feature encoding, and normalization techniques are used .the Hybrid Gray Level Co-occurrence Matrix Fast Fourier Transform (HGLCM-FFT) technique is used Feature extraction. The filter-based techniques such as symmetric uncertainty, IG, gain ratio, and CS are for used the Feature selection. The Hierarchical Gradient Boosted Isolation Forest (HGB-IF) algorithm is used for the classification .Trust authentication is used for security framework that functions across distributed cloud settings is provided a Distributed Adaptive Trust-Based Authentication (DAT-BA). By employing Particle Swarm Optimized Symmetrical Blowfish (PSOSB) algorithm is used for Encryption and decryption. The suggested work is implemented using OS Python – 3.9.6 , The performance of the suggested model are Attack Detection Rate ,False alarm rate, True positive rate, Network usage, CPU usage, Encryption time, Decryption time, Throughput.
Data Mining, Cloud computing, the Hybrid Gray Level Co-occurrence Matrix Fast Fourier Transform (HGLCM-FFT) algorithm, The Hierarchical Gradient Boosted Isolation Forest (HGB-IF) algorithm, Particle Swarm Optimized Symmetrical Blowfish (PSOSB) algorithm, Encryption time.
Joelle Mary Suji, Ananth S Mahadev, Nikita Elizabeth Chacko, Dr. Pradeep K V, School of Computer Science and Engineering, Vellore Institute of Technology, Chennai, Tamil Nadu
Software developers typically spend long stretches of time without any feedback about their mental states, which can cause fatigue, overload, and even burnout. In this paper, we introduce EMOTIA (Emotion-aware Multimodal Optimization for Task-adaptive Interaction and Assistance), an office robot and workstation cognitive wellness framework. EMOTIA consists of an ESP32-CAM edge computing device with multimodal perception, continual learning for customized affective baselines, cognitive load modeling taking context into account, explainable attribution, and episodic memory. Action units on the face, eye closure rate, head position, and prosody of speech are used to compute valence and arousal, which then drive preemptive non-intrusive intervention. A FastAPI backend is used to send telemetry using Server-Sent Events to the React web interface, while the FreeRTOS firmware draws expressive eyes on the OLED display. Our system met the requirements of all seven subsystems integration tests and REST/SSE endpoints verification. We present the architecture of our solution, the design of subsystems, and validation status, as well as remaining limitations for field deployment.
Affective computing, Human-robot interaction, Cognitive load estimation, Continual learning, Explainable AI
Salih Rashid Majeed1, Anthony Ibrahim2, Ezeddin Al Hakim3, and Tabassum Muqadam4, 1Plekhanov Russian University of Economics, Dubai, United Arab Emirates, 2Canadian University Dubai, Dubai, United Arab Emirates, 3Swibit, 4Plekhanov Russian University of Economics, Dubai, United Arab Emirates
Artificial intelligence (AI) is becoming central to the operation of modern power systems, giving rise to smart power management platforms that forecast demand, optimize dispatch, detect faults, and coordinate distributed energy resources in real time. This paper surveys recent advances in AI-driven smart power management and organizes the literature into a taxonomy of five technique families: classical machine learning, deep learning, reinforcement learning, federated or distributed learning, and agentic AI based on large language models (LLMs), and seven application domains: forecasting, demand response, fault diagnosis, digital-twin monitoring, edge intelligence, intelligent ERP, and industrial robotics. The survey presents a qualitative comparison of technique families across accuracy, real-time capability, scalability, and data-privacy dimensions.We then synthesize open challenges, including data quality, interpretability, cybersecurity, legacy integration, and agent oversight, and outline future directions such as physics-informed learning, privacy-preserving federated intelligence, and standardized benchmarking for AI-enabled smart power systems.
Artificial intelligence, Smart power management, Machine learning, Smart grid, Industrial robotics