DEV Community: Abdushshakur Sulaiman Abubakar

Aeglix Mind

Abdushshakur Sulaiman Abubakar — Sun, 02 Aug 2026 02:32:12 +0000

HAsH Guardian — Architecture & Vision

What It Is

HAsH Guardian is a free URL threat scanner built for everyday people in Africa and the Gulf. Paste any suspicious link — get a clear verdict in seconds. No account. No technical knowledge needed.

But the product is bigger than a scanner. It is a self-improving threat intelligence ecosystem that gets smarter every single day it runs.

Two Systems, One Product

HAsH Guardian — The Face

The product users interact with. Built on FastAPI with a clean Tailwind CSS interface. Deployed at aeglixmind.com.

Every scan runs the URL through seven global threat intelligence providers simultaneously — VirusTotal, Google Safe Browsing, PhishTank, ThreatFox, URLhaus, AbuseIPDB, and DNS Blocklist — combined with a custom ML model trained on over 500,000 real phishing and legitimate URLs. The result is a risk score from 0 to 1 with a clear LOW, MEDIUM, or HIGH verdict and a full explanation of why.

The Agent — The Brain

A separate autonomous cybersecurity agent running on dedicated infrastructure. When HAsH Guardian flags a URL as suspicious, it sends it to the agent through a secure encrypted tunnel. The agent does what a human analyst would do, investigates deeply, reasons about what it finds, and returns a structured result back to the scanner.

The agent runs three specialized open source AI models:

One for threat analysis
One for deep forensic investigation
One for writing new detection rules automatically

The agent is built with a skill-based architecture — every capability is its own isolated module. It runs 24 hours a day on permanently free infrastructure. No API limits, no monthly walls, nothing that can cut it off.

The Live Investigation Experience

When you scan a URL, you don't get an instant result. You watch the investigation happen in real time:

🔍 Analyzing domain registration history... ✅
🔍 Inspecting page content... ✅
🔍 Checking SSL certificate... ✅
🔍 Scanning for brand impersonation... ⚠️ Detected
🔍 Tracing redirect chain... ✅
🔍 Investigating IP reputation... ✅
🔍 Analyzing JavaScript behavior... ✅

Result: HIGH RISK

You see exactly why the verdict is what it is. You trust it because you watched it earn it. No other free tool does this.

Four Layers of Protection

HAsH Guardian doesn't just ask "is this site dangerous?" It asks four questions:

Layer 1 — Is this a direct threat?
Phishing, malware, known bad actors. Current scanner capability.

Layer 2 — Is this legitimate site safe to use?
Missing security headers, credentials submitted over unsecured connections, session hijacking risks, server information leaking to attackers.

Layer 3 — Is this site harvesting your data quietly?
Hidden tracking, fingerprinting scripts, data sent to unknown third parties, malvertising networks hiding in ads.

Layer 4 — Is this site hiding something time-delayed?
Malware embedded in ads or links, behavior that changes on return visits, redirects that only trigger on specific devices or locations.

A site can be completely legitimate and still fail Layers 2, 3, or 4. HAsH Guardian warns you either way — in plain human language, never technical jargon.

The Self-Improving Flywheel

Three independent data streams feed the ML model continuously:

Stream 1 — Agent discoveries
Every new threat pattern the agent finds gets verified and fed into the training pipeline through a dedicated endpoint.

Stream 2 — Global threat intelligence feeds
Live data from PhishTank, URLhaus, ThreatFox, CVE databases, and breach alert networks flows into the pipeline daily.

Stream 3 — Scanner scan history
Every scan HAsH Guardian processes generates real-world data that feeds back into training.

All three streams pass through a confidence score filter — an IoC must be confirmed across multiple independent sources before it touches the model. This prevents data poisoning from bad actors deliberately polluting threat feeds.

The result: the longer HAsH Guardian runs, the smarter it gets. The smarter it gets, the better it catches threats. The better it catches threats, the more people use it. The more people use it, the more data it generates. The cycle never stops.

The Mission

Western security tools are built for enterprises with budgets, IT teams, and technical expertise. Nobody is building intelligent, accessible, layered protection for everyday people in Africa and the Gulf — people who are increasingly online, increasingly targeted, and have nobody looking out for them.

HAsH Guardian exists to change that.

We don't consume intelligence. We produce it.

This is Aeglix. We are just getting started.

My Journey

Abdushshakur Sulaiman Abubakar — Thu, 23 Jul 2026 02:08:15 +0000

I built a URL threat detection system from scratch — here's what I learned from v0 to v2

A few months ago I posted here asking for advice and feedback on an early version of a phishing detector I was building. Today I'm sharing v2 — live, deployed, and significantly more powerful than where I started.
Here's the journey.

Where it started
The idea came from a real problem. People around me — friends, family, crypto community members — kept clicking suspicious links and getting scammed. There was no simple tool they could use to check a link before clicking. So I decided to build one.
v0 was basic. It worked but it was rough, limited, and I wasn't confident enough to share it widely. I actually hid it from my Facebook page because I was afraid of what people would say.
That fear wasted time. Lesson one.

What v2 became
v2 is a full threat detection system:
7 threat intelligence providers simultaneously — VirusTotal, Google Safe Browsing, PhishTank, ThreatFox, URLhaus, AbuseIPDB, DNS Blocklist
Custom ML model trained on 500,000+ real phishing and safe URLs
Risk score from 0 to 1 with full explanation
Deep scan mode for full site analysis
Bulk scanning for multiple URLs
Result caching for instant repeat scans
Results in under 20 seconds
What I learned

Ship even when it's not ready v0 taught me more than any tutorial. Real feedback exposed problems I never would have caught alone.
Provider redundancy matters When one provider is unavailable the system still works. Build for failure.
ML alone is not enough My ML model sometimes flags legitimate URLs. The lesson — combine multiple signals, don't trust any single one.
Transparency builds trust My DSA chairman reviewed the tool. His biggest concern wasn't detection — it was transparency. Users need to know what you check, how you score it, what you do with their data.

What's next
Privacy Policy and About page
Scoring methodology documentation
Private IP blocking to prevent abuse
Redis for persistent caching
Mobile app and browser extension
Try it
🔗 aeglixmind.com
Free. No login. No signup.
Feedback welcome — especially from security researchers and developers. Tell me what's wrong, what's missing, what could be better.

Built under Aeglix — cybersecurity tools for Africa. 🛡️

Support me with advice and feedback

Abdushshakur Sulaiman Abubakar — Thu, 22 Jan 2026 02:07:24 +0000

HAsh_Scanner — ML-Enhanced Phishing Detection System

I developed HAsh_Scanner, a web application that checks URLs in real time to detect phishing websites using machine learning and security checks.

Overview: Phishing attacks are one of the most common online threats, often using fake websites to steal passwords, bank details, or personal information. HAsh_Scanner is designed to help users verify links before visiting them, providing a clear risk assessment rather than a simple safe/unsafe label.

How It Works:

When a URL is submitted, the system performs multiple checks simultaneously.

URL Analysis: Examines subdomains, special characters, unusual encoding, and IP-based URLs. Detects patterns common in phishing links.

Domain Intelligence: Evaluates domain age, registration info, SSL certificate validity, and high-risk TLDs. New domains or unusual registrars are flagged.

Content Inspection: Searches for phishing keywords (verify, update, suspend, login) and common phishing page structures.

Machine Learning: Trained on over 156,000 real-world phishing URLs, the system identifies brand impersonation, suspicious paths, and subdomain abuse.

Risk Scoring: Aggregates all findings into a risk score from 0–100 to give users a nuanced view of potential threats.

Validation and Testing:

Tested on 51 legitimate websites, including major brands and banks: 0 false positives, average risk score 1.2.

Tested on 20 known phishing patterns: 70% detection rate, highest scores aligned with the most dangerous sites.

The system balances detection accuracy with minimal false positives to maintain trust.

Platform Security:

Custom middleware to handle requests securely.

Rate limiting (15 requests per minute, 100 per hour) and bot detection to prevent abuse.

Content Security Policy, XSS, and clickjacking protections.

HTTPS-only deployment ensures encrypted communications.

Technology Stack:

Backend: Python, Flask, Gunicorn

Machine Learning: Pattern recognition and heuristic analysis

Frontend: HTML5 / CSS3, responsive design

Deployment: Render with auto-deploy from GitHub, global CDN

Design Principles:

Minimize false positives

Explainable results rather than black-box decisions

Privacy-first approach: no URL storage or user tracking

Built for practical use in real-world scenarios, not only lab testing

Live Demo: https://hash-scanner-1.onrender.com

I welcome feedback from my mentors, developers and cyber security professionals to improve the system and expand it's capabilities.

Note: I am still a student and learning, so I may make mistakes. Your feedback is appreciated.