ScamShield Seniors

• 

  checking a single link or msg is suspicious/malicious/safe

• 

  checking batch of links or msg's are suspicious/malicious/safe

• 

  classifying links or msg's as suspicious/malicious/safe using ML

Inspiration

Scams targeting the older adults are increasing, and seniors often struggle to identify suspicious messages or links. Caregivers also need a simple way to quickly assess risk. This inspired ScamShield Seniors, a one-screen, accessible tool to help seniors and caregivers triage messages safely.

What it does

• Paste a message or URL → get a traffic-light verdict: Safe / Suspicious / Malicious.
  
• Shows top 3–5 reasons for the verdict in plain English.
  
• Provides copy-ready next steps and links to report scams.
  
• Optional batch mode: upload CSV of messages/URLs → receive scored CSV.
  
• Optional ML extension: combines heuristics with a pre-trained phishing model for improved accuracy.

How we built it

• Frontend: Streamlit for fast prototyping, large fonts, high-contrast buttons, and accessible design.
  
• Core logic: Rule-based heuristics analyzing keywords, urgency cues, and URL anomalies.
  
• ML extension: Optional pre-trained model for phishing detection to enhance accuracy.
  
• Extras: “Read aloud” toggle, color-blind safe icons, offline-first demo.

Challenges we ran into

• Balancing simplicity with accuracy as too many signals overwhelm seniors.
  
• Explaining reasoning clearly without cluttering the UI.
  
• Ensuring offline demo capability while also enabling optional ML features.
  
• Accessibility polish under a tight timeline.

Accomplishments that we're proud of

• Delivered a fully functional MVP in <4 hours.
  
• Accessible-first, one-screen design that is easy for seniors to use.
  
• Explainable heuristics that can be understood by both users and judges.
  
• Built a batch processing option and ML extension for future scalability.

What we learned

• Simplicity and explainability often matter more than complex models in MVPs.
  
• Accessibility improvements (fonts, contrast, keyboard navigation) dramatically improve usability.
  
• Rapid prototyping with Streamlit allows shipping working apps very quickly.
  
• Social impact is achievable with small, focused features.

What's next for ScamShield Seniors

• Browser extension for direct triage in messaging apps.
  
• WHOIS age and domain reputation signals for better detection.
  
• Lightweight ML models to improve accuracy without sacrificing explainability.
  
• Multilingual support and wider community distribution through libraries and senior centers.

Built With

• css3
• html
• machine-learning
• numpy
• python
• scikit-learn
• streamlit
• tldextract