RiskLens

“See the risk before you act.”

Comment

Inspiration

Digital financial scams have become a daily reality for millions of people, especially first-time digital users, students, and low-income individuals. Messages claiming urgent account issues, fake loan approvals, QR-code payment tricks, and impersonation scams exploit fear and urgency.

What inspired us most was noticing that existing tools either block content silently or show generic warnings without explaining why something is risky. Many victims realize the scam only after losing money.

We wanted to build a system that helps users pause, understand, and reason about risk before acting. This idea became RiskLens.

What it does

RiskLens is a multimodal AI-powered scam protection system that helps users verify suspicious financial interactions before taking action.

It allows users to:

Upload screenshots of SMS, WhatsApp messages, UPI requests, QR codes, or app screens

Receive a risk level (Low, Medium, or High)

Understand why something is risky through clear explanations

Learn the scam tactic being used (urgency, authority, reward, fear)

Get safe action recommendations instead of panic alerts

RiskLens focuses on understanding and prevention, not fear-based detection.

How we built it

RiskLens was designed as a reasoning-first system, not a chatbot.

Core components include:

Scam Guard for screenshot-based scam verification

A risk reasoning engine that analyzes urgency, identity mismatch, context, and visual cues

An explanation layer that converts AI reasoning into simple language

A safety guidance module that recommends responsible next steps

To unify multiple signals, we used a weighted risk scoring approach:

𝑅 𝑖 𝑠 𝑘

𝑆 𝑐 𝑜 𝑟

𝑒

𝑤 1 ⋅ 𝑈 + 𝑤 2 ⋅ 𝐼 + 𝑤 3 ⋅ 𝐶 + 𝑤 4 ⋅ 𝑉 Risk Score=w 1 ​

⋅U+w 2 ​

⋅I+w 3 ​

⋅C+w 4 ​

⋅V

Where:

𝑈 U represents urgency signals

𝐼 I represents identity mismatch

𝐶 C represents context inconsistency

𝑉 V represents visual deception cues

The score is translated into Low, Medium, or High risk, along with an explanation.

Challenges we ran into

Designing a calm experience We avoided fear-based language and focused on clarity and trust.

Explaining complex scam patterns simply The challenge was balancing simplicity with meaningful insight.

Responsible AI boundaries We chose not to auto-block payments and ensured the system advises rather than controls.

Accomplishments that we're proud of

Built a screenshot-first, multimodal scam verification feature

Designed explainable risk reasoning instead of black-box alerts

Created realistic scam scenarios users can learn from

Maintained a clean, product-grade UI

Focused on ethical and user-empowering AI design

What we learned

Most scam victims act under pressure, not lack of intelligence

Explanation builds trust more effectively than warnings

Visual context is critical for real-world scam detection

AI works best when it supports human judgment.

What's next for RiskLens

Planned improvements include:

Multilingual explanations

Region-specific scam pattern detection

Voice-based scam verification

Community-reported scam pattern learning

Greater personalization based on user experience level.

Built With

  • app.tsx
  • chatinterference
  • constants.tsx
  • css
  • geminiservice.ts
  • html
  • index.tst
  • javascript
  • metadatajson
  • scanguard
  • scenarioplanner
  • type.ts
  • ui
Share this project:

Updates