EdgePhishGuard
EdgePhishGuard is an Android application designed to detect phishing threats directly on the device, without sending user content to remote servers. It analyzes suspicious links, QR codes, text messages, and screenshots using on-device machine learning models (TensorFlow Lite), Bloom filters, and rule-based security heuristics. The app supports Spanish and English language detection, automatically routing content to the appropriate local model for inference. It also includes OCR capabilities to extract URLs from images captured by the camera or saved in the gallery, and can process content shared from other apps or notifications for real-time phishing alerts. The analysis pipeline normalizes input, identifies the language, runs local classifiers, checks URL reputation and domain patterns, and finally issues a warning only if the combined risk score exceeds a user-configurable threshold. EdgePhishGuard aims to provide transparent, explainable risk assessments, displaying the detected language, suspicious patterns, domain details, and a final security recommendation. By keeping all core analysis on-device, it addresses privacy concerns common in cloud-based security tools while still offering robust detection of smishing, quishing, and other mobile phishing vectors.
- Category
- Security Software
- Pricing
- Unknown
Problem solved
Mobile phishing attacks arrive through SMS, messaging apps, QR codes, and emails, often impersonating trusted brands to steal credentials or personal data. EdgePhishGuard solves this by enabling real-time, on-device detection of phishing links and suspicious content without exposing user data to external servers, thus preserving privacy and reducing reliance on constant internet connectivity.
Use cases
- Checking suspicious links received via SMS or messaging apps
- Analyzing QR codes for hidden phishing URLs
- Extracting and analyzing URLs from screenshots or photos using OCR
- Real-time alerts for phishing attempts in compatible notifications
- Evaluating text content for phishing patterns
- Configuring a personalized risk threshold to reduce false alarms
Audience
- Business Users
- Hobbyists
- Researchers
Pros
- All core analysis runs locally on the device, preserving privacy
- Combines ML models, Bloom filters, and rule-based checks for accuracy
- Supports OCR to extract links from images and QR codes
- Provides explainable risk scores with details on suspicious patterns
- Configurable alert threshold allows user to tune sensitivity
Cons
- Only available on Android (no iOS version)
- Reliance on local models may result in lower detection rates compared to cloud-based services
- Requires manual sharing of content for some scanning modes
- Limited language support (Spanish and English only)
- Potential for false positives/negatives due to local inference constraints
Official website: https://edgephishguard.web.app/
Canonical LambdaBase page: https://www.lambdabase.com/tools/edgephishguard