Did you receive a suspicious link?
Review the domain, urgency signals, and risk result before opening the page or entering credentials.
How to check a linkEdge protection
Anti-phishing defense built to run on Android, close to the user.
EdgePhishGuard combines local models, security rules, Bloom Filter checks, and contextual analysis to warn only when the configured risk threshold is exceeded.
Local
The primary analysis runs on the phone, not through a remote API.
AI
Local TFLite models evaluate text and URL risk signals.
OCR + QR
Extract text and links from the camera or local gallery.
Realtime
Alerts for compatible app notifications.
Real situations
Check suspicious content before you interact.
Start with the risk in front of you. These guides explain what to inspect and how EdgePhishGuard can provide an additional layer of protection.
Do you know where that QR code leads?
Learn how to spot quishing and analyze the URL hidden in a QR code from the camera or a saved image.
How to check a QR codeIs the link inside a screenshot?
Import a screenshot or use the camera to extract text and URLs with OCR before starting local analysis.
See the OCR flowKey features
Detect phishing where it actually appears.
The goal is not only to flag dangerous links, but to explain why the content is risky and what the user should do before opening it.
Local text and URL analysis
Paste, type, or share content from another app to estimate risk without sending the message or link to the Internet.
Language-routed local models
Detects Spanish or English and routes the text to the matching local model before combining risk signals.
On-device OCR and QR
Extract links from screenshots, photos, screens, or QR codes with a visual flow designed for fast review.
Notifications and shared content
Reviews compatible notifications and content shared from other apps to raise phishing alerts in real time.
Explainable risk
Shows final probability, detected language, suspicious pattern, blacklist signal, domain, and security recommendation.
Configurable threshold
The user controls when warnings appear; the initial recommended value is above 50% risk.
Technical pipeline
From suspicious input to a clear warning.
The flow combines language, structure, and reputation signals to reduce noise and warn only when there is enough evidence.
1
Input
Text, URL, notification, QR code, screenshot, or photo.
2
Extraction
Normalizes content and separates useful text from detected links.
3
Language
Identifies Spanish or English to choose the inference route.
4
Local model
Runs the right TFLite classifier directly on the device.
5
Bloom + rules
Checks reputation, URL patterns, and brand impersonation.
6
Alert
Warns only when final risk exceeds the configured threshold.
Android experience
Readable results for fast decisions.
The app shows risk, evidence, and recommended action without hiding the technical detail.
Warning with context
When risk is high, EdgePhishGuard shows the detected link, triggered signals, and a direct recommendation to avoid entering credentials or personal data.
OCR for screenshots and screens
The user can frame a message, page, or screenshot, select recognized text blocks, and extract only URLs or the full text.
Realtime alerts
Protection can detect suspicious content in compatible notifications and display a warning before the user opens the link.
Low risk is clear too
If the content does not exceed the threshold, the app says so clearly while keeping technical analysis available.
Official download
Available on Google Play
Install EdgePhishGuard from its official store listing.
Privacy by design
Sensitive content is analyzed on the device.
EdgePhishGuard is designed to review threats without turning the user's messages into content sent to an analysis server.
- No required account The app does not require account registration to analyze content locally.
- No message or screenshot uploads Texts, URLs, and screenshots reviewed by the user remain on the device during analysis.
- Local models Classification uses resources bundled with the app and local inference routes.
- Technical transparency The user can review final probability and the signals that caused each warning.