SpamGuard-AI

Inspiration

With the increasing number of digital scams—especially via SMS, emails, and messaging apps like WhatsApp—we wanted to create a tool that empowers people to quickly detect and respond to spam or scam messages. Most users don't have the technical know-how to analyze suspicious content. We aimed to fill this gap by using AI to make online communication safer and smarter.

What it does

SpamGuard-AI is an AI-powered web app that:

Detects if a given message is spam or not using a trained ML model.

Analyzes spam likelihood and gives a confidence score visually.

Supports voice-to-text input for catching spoken or WhatsApp voice message scams.

Offers multilingual input support for broader accessibility.

Suggests intelligent reply templates to handle spam politely or effectively.

Allows users to download the result as a PDF for evidence or reporting.

Includes a sleek light/dark mode UI toggle for a modern, accessible experience.

How we built it

Frontend: HTML, Tailwind CSS, JavaScript – with custom dark/light theme and responsive UI.

Backend: Python (Flask) – handles message input, ML model integration, and PDF generation.

ML Model: Logistic Regression/Naive Bayes trained on spam datasets (SMS, email).

Libraries: Scikit-learn, fpdf, speech recognition, langdetect, textblob for language processing.

Challenges we ran into

Handling voice input in a browser and converting it to accurate text required fine-tuning.

PDF export formatting without clutter was tricky with longer user inputs.

Making the UI mobile responsive and accessible while adding new features.

Ensuring the model performed well across multiple languages using lightweight NLP tools.

Accomplishments that we're proud of

Delivered a full-stack project with smart features in a short time.

Designed a modern, highly interactive UI that users loved.

Added multiple layers of functionality beyond detection, like voice input and reply suggestions.

Made security awareness more accessible and engaging to non-technical users.

What we learned

How to integrate speech-to-text and NLP models in real-time applications.

UI/UX design that adapts to user context, such as dark/light mode and mobile usage.

Designing intelligent interfaces that educate users instead of just showing predictions.

Deployment practices and optimizing ML inference in lightweight web environments.

What's next for SpamGuard-AI

Deploying the model as an API to support browser extensions for email/web-based spam filtering.

Adding a community reporting feature to share new spam messages for model retraining.

Integrating chatbot response automation with Telegram or WhatsApp bots.

Implementing explainable AI to show why a message was marked as spam, improving transparency.

Built With

• flask
• googletrans
• html
• javascript
• pdfkit
• python
• scikit-learn
• speech
• tailwind-css
• web