Computer Networks & Cybersecurity graduate with hands-on experience in network security, penetration testing, SOC analysis, IDS/IPS deployment, and AI-powered security solutions. Former trainee at Jordan's National Cyber Security Center (NCSC) and author of the IEEE paper "SemGuard: A Triple-Anchor Semantic Security Gateway for Multilingual Prompt Attack Detection in Large Language Models" (IEEE AEECT 2026, IEEE Xplore). Passionate about AI/LLM security, cyber defense, and adversarial machine learning research.
Introduced the first Triple-Anchor Semantic Threat Modeling framework for explainable, multilingual LLM prompt-attack detection, covering Arabic, Arabizi, and English. Built the first formally validated Arabic LLM Security Dataset (807 samples across 7 threat categories) using GPT-4o, Grok-4, and Llama 3.3 70B as independent judges, achieving a Fleiss' ฮบ of 0.839.
Keywords: LLM Security ยท Prompt Injection ยท AI Security ยท Arabic NLP ยท Explainable AI
Cite View on Hugging Face GitHubReframed the high inter-judge disagreement on LLM impersonation (98.2% in SemGuard) as a quantitative security signal rather than annotation noise. Introduced the Impersonation Ambiguity Index (IAI) to score inputs across four independent axes: Target Realism, Deceptive Intent, Consent/Context Boundedness, and Downstream Actionability, operationalized as an uncertainty-aware anchor for semantic gateways.
Keywords: LLM Security ยท Impersonation Detection ยท Role-Play Safety ยท Annotation Ambiguity ยท LLM-as-a-Judge
View Preprint (DOI) ResearchGateA deep technical breakdown of building a fully autonomous security pipeline. How I engineered two AI agents (GPT-4o & GPT-5.2) to attack, analyze, patch code, and re-test a vulnerable application in under 2 minutes with zero human intervention.
Read Technical Write-upAI-powered privacy policy analyzer built for Microsoft's Agents League Hackathon 2026 (Reasoning Agents track). Six specialized AI agents extract, reason, detect dark patterns, score readability, audit user rights, and benchmark policies against TikTok, Facebook & more โ covering GDPR/CCPA/PDPA/PIPEDA/LGPD/DPDPA. Built with Flask + Microsoft Foundry.
Read Technical Write-upููู ูุชูุงุนุจ ุงูู ูุงุฌู ูู ุจุนููู ุงููู ุงุฐุฌ ุงููุบููุฉุ ูู ุฃููู ู ูุงูุงุช ุงูุณูุณูุฉุ ูููู ุงูููุฏุณุฉ ุงูุนูุณูุฉ ุฎูู ูุฌู ุงุช ุงูู Jailbreaking ูุงูุฌุฐูุฑ ุงูุจููููุฉ ุงูุชู ุชุฌุนู ุงูุญููู ุงูุฃู ููุฉ ุงูุชูููุฏูุฉ ุชูุดู ูู ุตุฏูุง.
ุงูุฑุฃ ุงูู ูุงูI built PhishGuard: a phishing detection system that catches these attacks using nothing but classical algorithms from my Algorithms course. No machine learning. No training data. No internet connection required.
Read ArticleGAME DEVELOPER & MANAGER (2025 - 2026)
Cybersecurity Trainee (Mar-2026 - May-2026)
Contributed a merged pull request to the OWASP FIASSE project, adding 10 contrasting Python "With/Without" code examples covering SSEM attributes across Maintainability, Trustworthiness, and Reliability.
Translated SSEM security principles into practical secure/insecure implementation examples and closed OWASP FIASSE Issue #14.
View Merged PR OWASP FIASSE๐ SemGuard โ Creator of the first Arabic LLM Security Dataset, featured in the global Awesome AI Security list.
View Awesome List๐ ThreatLens โ Creator of an open-source threat intelligence CLI, officially featured in the global Awesome OSINT repository (27.5K+ โญ).
View Awesome OSINTWorld Islamic Sciences and Education University (WISE), Jordan โ (2022 โ 2026).
First-of-its-kind semantic security gateway protecting LLMs against prompt attacks across Arabic, Arabizi, and English, built on a Triple-Anchor Semantic Threat Modeling framework. Companion research accepted and presented at IEEE AEECT 2026 (IEEE Xplore : DOI: 10.1109/AEECT69724.2026.11657880 ).
Hugging Face GitHubAn autonomous, closed-loop multi-agent security lab using AutoGen, GPT-4o, and GPT-5.2 for automated Red Team exploitation and self-healing Blue Team code patching.
GitHubAI-powered privacy policy analyzer built for Microsoft's Agents League Hackathon 2026 (Reasoning Agents track). Six specialized AI agents extract, reason, detect dark patterns, score readability, audit user rights, and benchmark policies against TikTok, Facebook & more โ covering GDPR/CCPA/PDPA/PIPEDA/LGPD/DPDPA. Built with Flask + Microsoft Foundry.
GitHubFull attack/defense lifecycle across a 3-VM lab: exploited and patched SQLi, XSS, OS Command Injection, and File Upload vulnerabilities. Deployed Splunk HEC for real-time log forwarding, custom detection queries, and incident response.
GitHubOfficially featured in the "Awesome OSINT" repository (27.5K+ stars) as a recognized threat intelligence tool. Automates IOC enrichment (IP / Domain / Hash / CVE) across 6 APIs โ AbuseIPDB, VirusTotal, Shodan, OTX, URLScan, NVD โ with color-coded Excel, JSON, and CSV reports.
GitHubDeveloped the core computer vision system for "Rased," an automated road infrastructure monitoring app submitted for the Crown Prince Award. Trained a custom YOLOv11 model to detect road issues (e.g., potholes) and converted it to TFLite for real-time, low-resource mobile inference.
GitHubDesigned complex topologies simulating enterprise networks, configuring routing protocols (OSPF) and switching security.
Smart Campus CCNA LabsDeployed Snort 3 to monitor live network traffic with custom rules detecting and blocking real attack patterns.
GitHub (IDS) GitHub (IPS)๐ IEEE CTF 2025 Finalist (OSINT ยท Forensics ยท Web)
๐ Huawei ICT Competition โ National Qualifier, sole university representative
๐ IEEE IJSPC โ 1st Place in AI Track & Best Research Paper Ambassador Award
๐ AAU AI & Robotics โ Top 60 of 300+ finalists
๐ IEEE AEECT 2026 โ Author of an IEEE Conference Paper
IEEE Young Professionals (YP) & IEEE Student Branch Alumni.
Program Committee Member
Technical Instructor โ Game Development Workshop
IJSPC 2026 Ambassador - IEEE WISE Student Branch