AI-Powered Threat Detection

PHISHER ZERO

Four AI agents. Zero phishing attacks.
Real-time browser protection that thinks before you click.

<800ms Detection Speed
70+ Security Vendors
4 AI Agents
Scroll to explore
LIVE THREAT INTEL
⚠ Credential-harvesting page detected — blocked ⚡ Suspicious subdomain pattern identified ⚠ VirusTotal: 14/70 vendors flagged this URL ⚡ Urgency language detected in email body ⚠ IP-based hostname — no legitimate domain ⚡ URL shortener redirect chain detected ⚠ Impersonation of financial institution blocked ⚡ Homoglyph substitution in domain name ⚠ Credential-harvesting page detected — blocked ⚡ Suspicious subdomain pattern identified ⚠ VirusTotal: 14/70 vendors flagged this URL ⚡ Urgency language detected in email body

The Threat Is Real.
So Is The Defense.

Phishing attacks exploit human psychology — not software bugs. Millions of fake sites go live every year. Traditional blacklists miss zero-day domains for hours after launch. Phisher Zero changes the game with a multi-agent AI pipeline that reasons about deceptive intent, not just known patterns.

🎭

Social Engineering

Attackers impersonate banks, universities, and services. Our NLP Agent understands tone, urgency cues, and manipulation tactics — not just keywords.

🔗

Malicious URLs

Zero-day phishing domains bypass blacklists for hours. Our URL Agent applies 8 structural heuristics and cross-references 70+ global threat vendors.

🧠

AI Ensemble

No single model is enough. Four specialized agents pool their signals with a weighted ensemble and a red-flag override for confirmed threats.

💬

Plain-English Verdicts

Most tools give you a red icon and leave you guessing. Phisher Zero explains exactly why something is dangerous — in language anyone can understand.

Four Agents.
One Verdict.

🌐
Browser Input
Email body, page content, or URL submitted via Chrome extension popup
Agent 01
🤖

NLP Agent

Gemini 1.5 Flash analyzes tone, urgency, fear cues, impersonation patterns, and deceptive calls to action. Falls back to RoBERTa transformer locally.

Google Gemini 1.5 RoBERTa Fallback LangChain
Agent 02
🔍

URL Agent

Applies 8 structural heuristics (IP hostnames, subdomain depth, TLD anomalies, homoglyphs) then cross-references every link against VirusTotal v3's 70+ vendor database.

VirusTotal v3 Heuristics BeautifulSoup
Agent 03
⚖️

Ensemble Agent

Combines NLP and URL confidence scores through a weighted model. Red-flag override: a confirmed VirusTotal hit forces a Phishing verdict regardless of other signals.

Weighted Model Red-Flag Override 3-Tier Verdict
Agent 04
📢

Explainer Agent

Synthesizes flagged signals into a mentor-style plain-English explanation with specific findings and actionable safety tips — no technical knowledge required.

Gemini API Few-Shot Prompting XAI
🚨
PHISHING
Confidence: 97%
"This message uses artificial urgency and links to a domain flagged by 14 security vendors."
⚠️
SUSPICIOUS
Confidence: 61%
SAFE
Confidence: 94%
TECHNOLOGY STACK
Python 3.11
FastAPI
Google Gemini 1.5
LangChain
RoBERTa
VirusTotal v3
Chrome MV3
React 19
Docker
SQLite

See It In Action

Watch how Phisher Zero detects a phishing page in under 800 milliseconds — right inside your browser.

Live extension demonstration — Chrome browser

Everything You Need to Know

Full project documentation — from architecture and requirements to testing results and academic references.

Stop Phishing Before It Starts

Free. Open source. Works on any Chromium browser. No setup needed beyond loading the extension. Your first line of defence — powered by four AI agents working in concert.

Download Now — Add to Chrome
Compatible with Chrome, Edge, Brave, and all Chromium-based browsers · Manifest V3

Built By :

Final Year Project — Spring 2026 · National University of Computer and Emerging Sciences, Karachi

Project Supervisor Sir Minhal Raza
Co-Supervisors Miss Saeeda Kanwal & Sir Usama Antuley