AI-Powered Cyberbullying Detection and Prevention for Safer Digital Spaces
By Soumiya Marichelvan, Aitizaz Ali, Uzair Ali
Abstract
This research aims to address the growing issue of cyberbullying, particularly among youth, by developing an AI-powered system for real-time detection and prevention on digital platforms. The research focuses on understanding the various forms of cyberbullying and their detrimental effects on mental health. Using the Agile system development methodology, the research will follow an iterative process involving design, development, and testing phases to create a reliable detection tool. The system is intended to identify harmful online behaviour and alert parents or guardians promptly. Preliminary results suggest that the tool will feature an intuitive user interface and effective algorithms capable of detecting abusive interactions. This research is significant, as it proposes a technological solution to a pressing social issue, aiming to enhance the safety and well-being of young people online. Additionally, the research aligns with SDG 16 Peace, Justice, and Strong Institutions by contributing to the protection of children from online harm and promoting safer, more just digital environments.
Secure Multi-Party Computation for Cross-Institutional Query Federation without Raw Data Exposure in Healthcare Databases
By Haider H. Razzaq
Abstract
Cross-institutional healthcare research requires secure analysis of patient data distributed across hospitals while complying with the Health Insurance Portability and Accountability Act (HIPAA) and General Data Protection Regulation (GDPR). Existing federated query platforms restrict analyses to predefined aggregates and may expose institution-level statistics, whereas federated learning remains vulnerable to information leakage. This paper proposes a Secure Multi-Party Computation (SMPC) framework for privacy-preserving federated Structured Query Language (SQL) queries across multiple healthcare institutions without disclosing individual records or intermediate results. The framework introduces a federated query language and compiler that transforms queries into depth-bounded arithmetic circuits. It integrates verifiable secret sharing, Beaver-triple generation, and distributed key generation to support secure equality testing, comparison, record linkage, and aggregation. Protocol complexity and security guarantees are analyzed under semi-honest and malicious adversarial models, with malicious security approximately doubling communication requirements. An analytical cost model, calibrated using the SPDZ family of Multi-Party Computation (MPC) frameworks, evaluates scalability across 3–30 institutions and 1,000–500,000 patient records per institution. Results project that a representative cross-institutional linkage query requires less than one second and approximately 77 megabytes of communication for 15 institutions containing 10,000 records each. Four healthcare case studies demonstrate the framework’s potential applicability to rare-disease cohort discovery, drug-safety surveillance, and comparative-effectiveness research. The proposed framework provides a scalable approach to privacy-preserving healthcare analytics while supporting regulatory compliance and enabling secure collaborative research without centralizing sensitive patient information.
Solar-Powered Wireless Mesh Streetlight Monitoring System for Smart Cities
By Mahmood A. Al-Shareeda, Ahmed Thabet Abdul-Wahid, Atheer Hani Dinar, Hussein Mohammed Mustafa, Wael Salman Jabbar
Abstract
Urban road lighting is a large-scale municipal controllable energy source with enormous room for optimization. At present, most already deployed street lamps still operate on fixed schedules, maintaining uniform full brightness across all time periods regardless of actual lighting demand. The solar-powered smart street lamp platform developed in this project is adapted for smart city scenarios, integrates core functional modules including renewable energy collection and adaptive dimming, and is equipped with an ESP32 controller, light sensors, and motion sensing hardware. It enables intelligent operation, comprehensively improving the efficiency, safety, and maintain-ability of street lamps. The core objective of the solar-powered connected smart streetlight system designed in this study is to leverage a wireless network architecture and management control dashboard to achieve synchronized reporting of the status, power level, and charging data of large-scale streetlight deployments, as well as fault alarms, to support remote management and control for operation and maintenance personnel. We built a prototype and completed three tests: adaptive brightness control reduced energy consumption from 288Wh to 108Wh, marking a 62.5% energy saving; motion-triggered lighting registered an average response time of 180ms and a detection success rate of 97.9%; the Blynk-based remote monitoring function recorded an average cloud latency of 245ms and a connection availability rate of 99.7%. This system can substantially cut energy consumption and improve the visibility and response efficiency of operation and maintenance efforts for smart cities.
Blockchain-Enabled Zero Trust Architecture for Securing Cloud Computing based on Cybersecurity Risk Auditing Framework
By Mohammed Almaiah, Santosh Reddy Addula
Abstract
Traditional cloud security controls tend to treat access, vulnerability exposure, and audit integrity as distinct functions, reducing the ability to be aware of risks on an ongoing basis. In this study, the authors present a blockchain-based zero-trust risk auditing framework for adaptive cloud protection (ACP) called BAZTRA. BAZTRA is a combination of a GBDT-based network-risk estimator and identity, device-posture, provenance, vulnerability, and compliance evidence. The CICIoT2023 corpus, controlled contextual replay, and repeated ten-seed experiments were used to evaluate the following techniques: availability-aware fusion, temporal smoothing, four-state enforcement, EPSS-KEV prioritization, hash chaining, and Merkle-based audit anchoring. Statistical significance was assessed using Friedman and Holm-corrected Wilcoxon tests across repetitions. The selected network-risk estimator achieved 78.96% accuracy, 55.01% macro-F1, 0.964 ROC-AUC, and 0.718 MCC. The complete BAZTRA framework produced 89.44% decision macro-F1, reduced unsafe access to 0.67%, and limited false denial to 0.03%. Ablation confirmed that adaptive enforcement, network evidence, provenance analysis, and availability-aware normalization were the most influential components. Merkle batching reduced effective ledger payload by 98.21% while preserving detection of modification, deletion, reordering, and replay attempts. BAZTRA provides a unified approach to continuous access assessment, vulnerability prioritization, and accountable cloud auditing. Future work will validate the framework on complete multi-source telemetry and a physical multi-peer Hyperledger Fabric deployment.
An Explainable AI Framework for Real-Time Cyberattack Detection and Risk Assessment
By Vugar Abdullayev, Alex Khang, Nazila Ragimova
Abstract
As cyberattacks become more sophisticated and more common, detection systems must be not only accurate but also quick and easy to interpret. Although deep learning models have proven to be very effective in detecting intrusion and malware, their black-box nature has hindered their adoption in security operations centers (SOCs), where analysts need to explain and take action on alerts in short timeframes. To offer real-time cyberattack detection with quantitative risk assessment, this paper presents an Explainable AI (XAI) framework that integrates a lightweight hybrid deep learning detector with post-hoc and intrinsic explainability methods. The framework combines a Convolutional-Bidirectional Gated Recurrent Unit (CNN-BiGRU) architecture for spatio-temporal traffic representation with an attention layer that highlights salient time steps and integrates SHAP (Shapley Additive explanations) and LIME (Local Interpretable Model-agnostic Explanations) to generate human-readable justifications for every prediction. A dynamic risk-scoring engine then translates detection confidence, feature-level severity, and asset criticality to a normalized risk index that enables prioritized incident response. The framework is tested on three benchmark intrusion detection datasets (CICIDS2017, NSL-KDD, and UNSW-NB15) and a real-time traffic stream generated synthetically. Experimental results reveal that the proposed model achieves a detection accuracy of 98.6%, achieves an F1-score of 0.981, and the average latency of inference is 14.2 milliseconds per flow, and explanation fidelity, based on the two metrics of deletion/insertion and rank-correlation stability, demonstrates that the explanations generated are faithful to the model's behavior. It has been shown that the proposed hybrid framework is suitable for application in real-time network security environments with limited resources, as it outperforms the standalone CNN, LSTM, Random Forest, and CNN-LSTM baselines in terms of detection performance and retains interpretability with only a minor computational cost. It also demonstrates the applicability of the framework for security operations workflows by means of a comprehensive ablation study, ROC and confusion-matrix analysis, and a comparison with the dynamic risk index and static CVSS-style scoring.
E-QSafe Energy-Aware Quantum-Safe Encryption Scheme for Power-Controlled Internet of Nano Things
By Mahmood A. Al-Shareeda, May Adnan Faleh, Mohammed AminAlmaiah
Abstract
The Internet of Nano Things enables nanoscale devices to perform critical operations in healthcare, industry, and environmental monitoring. Nevertheless, ultra-small systems encounter tremendous challenges, including energy limitation, computation weakness, and exposure to quantum-era threats. No conventional cryptographic solution can maintain the effectiveness of both efficiency and post-quantum resilience under the above severe constraints. This paper designs E-QSafe, an Energy-Aware Quantum-Safe Encryption Scheme, to address the concern, which is featured by integrating an adaptive power control method with lattice-based Identity-Based Encryption. More specifically, the E-QSafe dynamically tunes the lattice dimensionality and Gaussian sampling variance based on the device’s residual presenting ratio to secure the communication with minimal energy dissipation. The mathematical optimization model is developed to adjust the energy consumption, delay, and strength during the encryption at various power levels. Extensive experiments and analysis validate E-QSafe’s security against quantum, replay, and side-channel attacks as well. The evaluation results show that the E-QSafe obtains a 35% reduction in total energy dissipation, reduces the communication cost by 30%, and enhances the throughput compared with conventional IoNT schemes. In short, E-QSafe effectively realizes quantum-resilient and energy-sustainable secure IoNT for nanonetwork computing in the next age.
Secure-Aggregation Federated Graph Attention Networks with SHAP-Based Edge Attribution for Advanced Persistent Threat Detection
By Ahmed Fakhri Abdullah, Ahmed Sedeeq Baker Al-Doori
Abstract
Advanced Persistent Threats (APTs) pose significant cybersecurity challenges due to their stealthy, prolonged activities and the limited visibility of malicious behaviour in host provenance data. Existing detection approaches often require centralized access to sensitive logs and provide limited interpretability, raising concerns about data confidentiality and analyst trust. To address these challenges, this study proposes a privacy-preserving and explainable federated learning framework for APT detection using provenance graphs. The proposed framework integrates edge-typed Graph Attention Networks (GATv2), self-supervised masked-feature learning, and federated prototype-based anomaly detection to identify suspicious activities without requiring labeled attack data during training. To preserve confidentiality, a dropout-resilient, Bonawitz-style secure aggregation protocol is employed to protect model updates and shared statistics, including normalization parameters, clustering information, and calibration distributions. Furthermore, edge-level Shapley value attribution is incorporated to provide interpretable explanations of detected anomalies and support security analysts during incident investigation. The framework is evaluated using the DARPA Transparent Computing Engagement 3 benchmark, specifically the THEIA and CADETS datasets, under a simulated four-client federated setting. Experimental results across three random seeds demonstrate strong detection performance, achieving Area Under the Receiver Operating Characteristic Curve (AUROC) values of 0.995 and 0.971 and Area Under the Precision–Recall Curve (AUPRC) values of 0.870 and 0.561 for THEIA and CADETS, respectively. The proposed secure aggregation mechanism maintains detection performance comparable to conventional federated averaging, introduces approximately 0.1% additional training time on THEIA, and reduces cross-round client-linkage attack accuracy from 85.6% to chance-level performance. Behaviour-level analysis demonstrates complete detection of scan-like attack activities at a calibrated 95% threshold, while other malicious activities exhibit threshold-dependent recall. The explainability analysis confirms the completeness of Shapley attributions and convergence of the highest-ranked contributing edges within 25 permutations, although their empirical fidelity is comparable to less computationally expensive explanation methods. Overall, the findings demonstrate the potential of combining secure federated learning, provenance-graph representation learning, and explainable AI for confidentiality-preserving APT detection, while highlighting important limitations related to threshold sensitivity, explanation fidelity, and semi-honest security assumptions.