Fraud Recovery Assistant – Project Summary
Your Name: ___________
Project Type: Final Year Computer Science
Technology: Python, Streamlit, Pillow
1. Problem Solved
Scam victims panic and don’t know what to do next. No single tool asks them the right
questions and gives immediate, step-by-step recovery steps with phone numbers.
2. Solution
A conversational web app that:
• Guides the user through 5 simple questions (scam type, platform, amount lost,
date, description).
• Instantly displays a personalised recovery plan with contacts for banks, credit
bureaus, gift card companies, crypto exchanges, and law enforcement.
• Allows image upload (screenshots as evidence).
• Lets users download a text report (checklist & timeline).
• Supports English, Hindi, and Spanish (easily extendable).
3. Key Features
• 12 scam types covered (bank, crypto, gift card, identity theft, tech support,
romance, shopping, job, rental, phishing, etc.)
• Platform-specific help (Facebook, PayPal, WhatsApp, email, phone, etc.)
• Real contact numbers (Equifax, Amazon gift card hotline, Coinbase support,
FTC, IC3, etc.)
• Warning about recovery scammers
• Clean chat interface (Streamlit)
4. Technology Stack
Component Tool/Library
Frontend/UI Streamlit
Component Tool/Library
Logic Python 3.10+
Images Pillow (PIL)
Language data Python dictionaries
Report Text file (download)
Deployment Localhost / Streamlit Cloud
5. How It Works (Flow)
1. User selects language.
2. Assistant asks: What scam type? → User clicks a button.
3. Which platform? → User clicks.
4. How much money lost? → User types.
5. When did it happen? → Date picker.
6. Describe what happened? (optional) → User types.
7. App generates recovery plan with contacts and actions.
8. User can upload screenshot and download report.
6. Sample Output (for Bank Transfer Scam)
text
IMMEDIATE ACTIONS
1. Call your bank using number on back of card.
2. Freeze credit with Equifax (1-888-349-9960), Experian (1-888-397-3742), TransUnion
(1-888-909-8872).
3. File FTC report: [Link]
4. Change passwords and enable 2FA.
Warning: Recovery scammers will contact you – ignore them.
7. What Makes This Project Special
• Victim-centric – helps after the scam, not just prevention.
• Actionable – gives exact phone numbers, not just advice.
• Multi-language – accessible to non-English speakers.
• Evidence ready – image upload + downloadable report for police/insurance.
8. Future Enhancements (Optional)
• Machine learning to auto-detect scam type from user description.
• Real-time URL / QR code scanning.
• Integration with police/FTC APIs for direct filing.
• Voice interface.
9. How to Run
bash
pip install streamlit pillow
streamlit run final_app.py
10. Expected Professor Q&A (Short Version)
Q: Why Streamlit?
A: Fast, built-in chat & file upload, no frontend coding.
Q: How do you ensure phone numbers are correct?
A: Verified from official sources at development time; in production they would come
from a live API.
Q: Is there any machine learning?
A: Not in this version – rule-based is reliable for recovery steps. But we have a plan to
add ML for scam classification.
Q: How is this different from the FTC website?
A: FTC is static; our app converses with the victim, asks questions, and delivers a
personalised, platform-specific plan with direct numbers.
Q: What about user privacy?
A: No database; all data stays in browser session and is deleted on refresh.
Use Case Diagram
Actors:
• User (scam victim)
Use Cases:
1. Select language
2. Identify scam type
3. Provide platform details
4. Enter amount lost
5. Provide date & description
6. View recovery plan
7. Upload image evidence
8. Download report
9. Ask follow-up questions
How to Draw ([Link]):
• Actor (stick figure) labelled “Victim”.
• Oval use cases around it.
• Connect actor to each use case with a line.
• No system boundary needed for simplicity.
Alternative: I can describe the diagram in text, but you can easily create it using the
above list.
Slide 7: Sequence Diagram (Normal Flow)
Actors: User, System (Streamlit App)
1. User → Opens app → System displays language selection.
2. User → Selects language → System loads translations.
3. User → Clicks scam type → System stores answer, asks platform.
4. User → Selects platform → System stores answer, asks amount.
5. User → Enters amount → System stores, asks date.
6. User → Picks date → System stores, asks description (optional).
7. User → Enters description (or skips) → System generates recovery plan.
8. User → Uploads image → System displays image and placeholder analysis.
9. User → Clicks download → System creates text report for download.
How to draw: Lifelines (User, System), arrows for each step, self-arrows for internal
processing.
Slide 8: System Architecture Diagram
text
┌─────────────────────────────────────────────────┐
│ User Browser │
└─────────────────┬───────────────────────────────┘
│ HTTP
┌─────────────────────────────────────────────────┐
│ Streamlit Server │
│ ┌─────────────────────────────────────────┐ │
│ │ Language Selector Module │ │
│ └─────────────────────────────────────────┘ │
│ ┌─────────────────────────────────────────┐ │
│ │ Scam Type & Platform Questionnaire │ │
│ └─────────────────────────────────────────┘ │
│ ┌─────────────────────────────────────────┐ │
│ │ Recovery Plan Generator (Dictionary) │ │
│ └─────────────────────────────────────────┘ │
│ ┌─────────────────────────────────────────┐ │
│ │ Image Upload Handler (Pillow) │ │
│ └─────────────────────────────────────────┘ │
│ ┌─────────────────────────────────────────┐ │
│ │ Report Generator (Text File) │ │
│ └─────────────────────────────────────────┘ │
│ ┌─────────────────────────────────────────┐ │
│ │ Chat Interaction Handler │ │
│ └─────────────────────────────────────────┘ │
└─────────────────────────────────────────────────┘
Explanation: All components run inside the same Python process. No external
database – session state stores user answers temporarily.
Slide 9: Activity Diagram (User Journey)
Start → Language Selection → Scam Type Selection → Platform Selection → Amount Entry
→ Date Entry → Description (optional) → Decision: All info collected? → Yes → Generate
Recovery Plan → Show plan, contacts, warnings → User can upload image / download
report → End
How to draw:
• Rounded rectangles for activities.
• Diamond for decision.
• Arrows showing flow.
• Swimlanes optional.
Slide 10: Class Diagram (Simplified)
Only three main “classes” in the code (Python dictionaries + session state):
• LanguageManager
o languages: dict
o translations: dict
o get_text(lang, key): string
• RecoveryEngine
o recovery_actions: dict (scam_type → actions)
o get_immediate_actions(scam_type): dict
• SessionState (Streamlit built-in)
o lang: string
o scam_answers: dict
o current_step: string
o messages: list
Relationships: SessionState uses LanguageManager and RecoveryEngine.
Slide 11: Sample User Interaction (Screenshot or Text)
User selects: English → Bank transfer scam → Bank platform → Amount $2000 → Date
2026-04-01 → Description “Someone called pretending to be my bank.”
Recovery Plan (excerpt):
text
IMMEDIATE ACTIONS
1. Call your bank using number on back of card.
2. Freeze credit: Equifax 1-888-349-9960, Experian 1-888-397-3742, TransUnion 1-888-
909-8872.
3. File FTC report: [Link]
4. Change passwords and enable 2FA.
Warning: Recovery scammers will contact you – ignore them.
Visual: Include a mockup of the chat window showing assistant messages.
Slide 12: Demonstration / Screenshots
• Screenshot 1: Language selection screen
• Screenshot 2: Scam type buttons
• Screenshot 3: Platform selection
• Screenshot 4: Recovery plan with contact numbers
• Screenshot 5: Image upload and download report button
(You can capture these from your running app.)
Slide 13: Testing & Validation
• Tested with 5 scam scenarios: bank, crypto, gift card, identity theft, tech
support.
• Verified contact numbers from official sources (FTC, Equifax, Amazon, etc.).
• Checked language translations with native speakers (basic check).
• Edge cases: User skips amount, description; no image uploaded – app still
works.
Result: App provides correct, actionable steps for each scam type.
Slide 14: Future Enhancements
1. Machine Learning Scam Classifier
o Train on real scam descriptions to auto-detect scam type from free text.
2. Real-time URL / QR scanning
o Integrate VirusTotal API to check malicious links.
3. Image Forensics
o Detect deepfakes or altered screenshots.
4. Direct Filing to Authorities
o Submit reports to FTC/IC3 via their APIs.
5. Voice Interface
o Speech-to-text for easier input.
6. Mobile App
o Convert to Android/iOS using BeeWare or Kivy.
Slide 15: Challenges & Lessons Learned
Challenge Solution
Managing multi-language text without external
Used nested Python dictionaries; easily extendable.
libraries
Keeping track of user progress across multiple
Leveraged Streamlit’s session_state.
steps
Handling image upload without permanent Streamlit’s file_uploader keeps image in memory; not
storage saved.
Researching correct contact numbers Consulted official FTC, Equifax, Amazon gift card pages.
Lesson: Simple, deterministic logic is often more reliable than complex AI for critical
recovery steps.
Slide 16: Conclusion
• What we built: A compassionate, actionable, multi-language fraud recovery
assistant.
• Impact: Helps victims immediately after a scam – fills a gap left by
prevention-only tools.
• Uniqueness: Step-by-step interview + platform-specific contacts + evidence
upload + downloadable report.
• Future ready: Can integrate ML, APIs, and mobile deployment.
Final Statement: “This project proves that technology can be a lifeline, not just a
detector.”
Slide 17: Questions & Answers
Prepare these answers (already provided in earlier documentation).
• Q: Why Streamlit?
• Q: How do you ensure contact numbers are correct?
• Q: Is there any database? Privacy?
• Q: How is this different from FTC website?
• Q: Can it detect new scams?
• Q: What would you improve given more time?
Slide 18: References & Thank You
• FTC: [Link]
• IC3: [Link]
• Equifax, Experian, TransUnion
• Streamlit Documentation
• Python Software Foundation
Thank you!
Any questions?
How to Create the Diagrams (Free Tools)
1. [Link] (now [Link]) – completely free, online or desktop.
o Go to [Link].
o Choose “Create New Diagram”.
o Use shapes: Actor (use case), Lifeline (sequence), Rectangle (class), etc.
o Export as PNG or PDF and insert into your presentation.
2. Lucidchart – free tier with limited shapes, good enough for this project.
3. PowerPoint / Google Slides – you can draw basic shapes manually.
Quick diagram templates (text-based you can copy into [Link]):
Use Case Diagram (text format for [Link])
text
Victim --> (Select language)
Victim --> (Identify scam type)
Victim --> (Provide platform)
Victim --> (Enter amount lost)
Victim --> (Enter date & description)
Victim --> (View recovery plan)
Victim --> (Upload image)
Victim --> (Download report)
Victim --> (Ask follow-up questions)
Sequence Diagram (Mermaid format – can be rendered in some markdown viewers)
(You can use Mermaid Live Editor to convert this to an image.)