Visual phishing detection engine for East African financial and telco brands. Screenshots suspected URLs, encodes with CLIP, and compares against a reference index of legitimate brand interfaces. A visual layer for OCTIO's threat intelligence pipeline.
# visual-threat-intel
Visual phishing detection engine for East African financial and telco brands.
Screenshots suspected URLs, encodes with CLIP, and compares against a
reference index of legitimate brand interfaces. A visual layer for OCTIO's
threat intelligence pipeline.
## Protected brands
Safaricom, M-Pesa, Equity Bank, KCB, Co-operative Bank, Airtel Kenya,
Standard Chartered Kenya, DTB, Absa Kenya, Flutterwave.
## How it works
## Setup
```bash
pip install -r requirements.txt
playwright install chromium
# capture reference images from official brand sites
python3 -c "
from visual_threat_intel.phishing.brands import BRANDS
from visual_threat_intel.phishing.screenshotter import capture_reference_images
capture_reference_images(BRANDS)
"
# build the brand reference index
python3 -c "
from visual_threat_intel.phishing.index import PhishingIndex
PhishingIndex().build()
"
```
## Usage
```python
from visual_threat_intel.phishing.detector import PhishingDetector
detector = PhishingDetector()
result = detector.scan("
safaricom-ke-login.suspicio…")
print(result)
# {
# "phishing": True,
# "matched_brand": "safaricom",
# "matched_brand_name": "Safaricom",
# "similarity": 0.91,
# "official_url": "
safaricom.co.ke",
# "reason": "domain_mismatch"
# }
```
## Part of OCTIO-Labs
This module is the visual intelligence layer of the OCTIO threat detection
pipeline. Text-based URL classification lives in
OCTIO-Labs/OCTIO.