Paper Details

Name: DIPAK VIJAY RAJPUT , SUNIL VERMA
Country: India

Year: 2025
Volume: Volume-12, Issue-2 (July-December)
Page Number: 527-543
DOI: https://doi.org/10.5281/zenodo.21428901

ABSTRACT
The rapid expansion of digital social systems, including social media platforms, cloud services, Internet of Things (IoT) networks, and online collaborative environments, has significantly increased the complexity and frequency of cybersecurity threats. Traditional cybersecurity solutions often rely on static rule-based mechanisms and standalone machine learning models, which suffer from high false-positive rates, limited adaptability, delayed incident response, and inadequate risk prioritization in dynamic environments. To address these limitations, this study proposes a Sensitivity-Aware Hybrid Intelligent Framework (SAHIF), a novel decision-support framework for real-time cybersecurity supervision in social systems. The proposed framework integrates Convolutional Neural Networks (CNN), Bi-Long Short-Term Memory (Bi-LSTM), Transformer networks, Random Forest, Attention Mechanism, and Decision Fusion to enhance cyber threat detection and classification. A multi-layer Sensitivity Analysis Module computes the User Sensitivity Score (USS), Data Sensitivity Score (DSS), Asset Sensitivity Score (ASS), Context Sensitivity Score (CSS), and Overall Sensitivity Index (OSI) to prioritize cybersecurity risks intelligently. The framework incorporates data collection, preprocessing, feature engineering, adaptive risk assessment, automated decision support, incident response, and continuous learning to ensure resilient cybersecurity supervision. Experimental evaluation is conducted using benchmark datasets, including CICIDS2017, CICIDS2018, UNSW-NB15, NSL-KDD, TON-IoT, Bot-IoT, and the Twitter Bot Dataset. Performance is assessed using Accuracy, Precision, Recall, F1-score, ROC-AUC, Detection Rate, False Positive Rate, Matthews Correlation Coefficient (MCC), Processing Latency, and Throughput. The proposed framework demonstrates superior detection accuracy, improved real-time decision-making, reduced false alarms, and enhanced adaptability compared with existing cybersecurity approaches, making it suitable for deployment in modern, dynamic social systems.

Keywords: Cybersecurity, Artificial Intelligence, Hybrid Intelligence, Sensitivity Analysis, Real-Time Monitoring, Social Systems, Threat Detection, Decision Support.