BharatShield AI

About BharatShield AI Inspiration

Cyber scams are becoming increasingly sophisticated, while many people still struggle to identify whether a message, link, QR code, or online request is safe. Phishing messages often create urgency, impersonate trusted organizations, or use convincing language to trick users into sharing sensitive information.

We were inspired to build BharatShield AI to provide an easy-to-use AI-powered safety layer that helps everyday users understand digital threats before they act on them.

What We Built BharatShield AI is a multimodal AI cybersecurity assistant that analyzes suspicious messages, URLs, QR codes, and screenshots to identify potential scams and explain why they are risky.

Instead of simply displaying “Scam” or “Safe,” our system provides an understandable risk assessment with the indicators that influenced the decision.

For example: High Risk — 94% Threat: Financial Phishing Detected: Urgency, impersonation, suspicious URL, credential request Recommendation: Do not click the link or share sensitive information.

Our goal is to make cybersecurity understandable even for users without technical knowledge.

How We Built It The system follows a multimodal pipeline:

User Input ↓ Message / URL / QR / Screenshot ↓ Text Extraction & Feature Analysis ↓ AI Threat Detection ↓ Risk Classification ↓ Explainable AI ↓ Risk Score + Threat Type + Safety Recommendation

For screenshots, OCR is used to extract relevant text before sending it through the threat-analysis pipeline. URLs are analyzed for suspicious characteristics, while QR codes are decoded and their destinations are evaluated for potential risks.

The system combines Python, machine learning, NLP, OCR, computer vision, React.js, and a backend API to create a complete end-to-end application.

What We Learned During development, we learned that building an effective AI application involves much more than training a model. We gained practical experience in:

Multimodal AI and combining different input types NLP-based scam and phishing detection OCR and image processing URL and threat-feature analysis Explainable AI and risk scoring Building APIs and connecting AI models to a web application Designing an intuitive cybersecurity interface Evaluating AI predictions and handling uncertain results Deploying an AI application as a usable product

One important lesson was that explainability matters. Users are more likely to trust and act on an AI warning when the system clearly explains why something was flagged.

Challenges We Faced One of our biggest challenges was handling different types of cybersecurity inputs within a single system. Messages, URLs, screenshots, and QR codes require different processing techniques.

We also faced challenges in reducing false positives. A legitimate message can sometimes contain urgency or links, while a scam can be carefully written to appear genuine. Therefore, we designed the system to consider multiple indicators instead of relying on a single keyword or rule.

Another challenge was presenting complex AI results in a simple way. We focused on converting technical model outputs into clear information such as:

Risk → Threat Type → Reasons → Recommended Action

Real-World Impact BharatShield AI is designed as a preventive cybersecurity tool for students, families, senior citizens, and everyday internet users.

Our vision is simple: People should not need to be cybersecurity experts to stay safe online.

By combining AI-powered detection with clear explanations and actionable recommendations, BharatShield AI aims to help users detect scams, understand threats, and make safer digital decisions before it is too late.