Privacy Policy

1. Overview and Application Identity

AI Phishing Shield (officially titled the AI Phishing Email Detection System; hereinafter “the Application”, “we”, “our”, or “the Service”), accessible at https://www.emailphisingdetector.online, is a cybersecurity monitoring platform designed to help individuals and security analysts identify deceptive emails, spoofed sender domains, credential-harvesting links, and Business Email Compromise (BEC) fraud.

2. Google API Services User Data Policy & Limited Use Disclosure

In strict compliance with Google's Limited Use standards:

• No Generalized AI/LLM Training: User email content obtained via Google APIs is never transferred, exported, or used to develop, train, retrain, or fine-tune generalized artificial intelligence, machine learning, or commercial foundation models (including Large Language Models).
• No Commercial Reselling or Advertising: Google user data is never sold, leased, rented, or transferred to third-party data brokers, advertising networks, or marketing aggregators.
• Strict Purpose Limitation: Information received via Google APIs is used solely to provide and improve user-facing email phishing analysis, threat detection, and forensic scoring.
• No Human Review: No human being reads your email communications unless:
  1. You provide explicit affirmative consent for a specific message to troubleshoot a false-positive detection;
  2. Necessary for security purposes (such as investigating system abuse); or
  3. Required to comply with applicable statutory laws or judicial processes.

3. Google OAuth Scopes & Purpose of Access

The Service connects to Google APIs solely upon your voluntary affirmative action. We request only the minimum necessary permissions under the principle of least privilege:

4. Information Collected, Processed & Stored

The application processes only the data strictly required for threat detection:

A. User Identity & Profile Data

• Google Account Email: Used as your unique tenant identifier.
• Profile Display Name: Displayed in the top navigation header.
• Google User ID: Internal alphanumeric ID used to enforce cross-user data isolation.

B. Google OAuth Credentials

• Access & Refresh Tokens: Received from Google's token endpoint upon successful OAuth consent.
• Storage & Encryption: Both access and refresh tokens are encrypted at rest using symmetric Fernet encryption (AES-128-CBC with SHA-256 HMAC). Plaintext tokens are never stored in the database and are never sent to browser local storage.

C. Email Message Metadata & Content

• Header Metadata: Message ID, Subject, Sender address (From), Recipient address (To), and Date/Time timestamp.
• Message Body Snippets: Plaintext body text and HTML markup extracted to scan embedded URLs, domain reputations, urgency patterns, and suspicious formatting.

D. Forensic Security Artifacts & Risk Scores

• Extracted URLs: Hyperlinks found within scanned emails, inspected for raw IP addresses, typosquatting, and suspicious top-level domains (.xyz, .top, .buzz, etc.).
• Machine Learning Probability: A statistical probability score (0.0 to 1.0) generated by our local Scikit-learn TF-IDF vectorizer and Logistic Regression classifier.
• Heuristic Scores & Flags: Specific detected threat indicators (e.g., “Urgency tactics detected”, “Sender domain mismatch”, “Credential harvesting link”).
• Composite Risk Score: An aggregated score from 0 to 100 categorizing messages as SAFE (0–39), SUSPICIOUS (40–69), or PHISHING (70–100).

E. Manual Sandbox Submissions

• Any custom email subjects, senders, and body texts manually submitted into the Sandbox Scanner for on-demand analysis.

F. Session & Authentication Tokens

• Cryptographically signed JSON Web Tokens (HS256 JWT) stored in browser local storage (phish_auth_token, phish_user_id) to maintain authenticated sessions across page reloads.

5. Storage Location, Retention Schedule & Data Deletion

A. Storage Location

In production, all account credentials (encrypted), email scan records, threat notifications, and forensic analysis results are stored in a persistent MongoDB database hosted in the cloud. In local development mode, a resilient local JSON storage file is utilized.

B. Cross-User Data Isolation

All stored records are partitioned and indexed strictly by the authenticated user_id. Every database query enforces user ownership verification. No user can view, search, modify, or delete another user's email records.

C. Retention Period

Scanned email records and threat logs remain stored in the user's isolated workspace until manually deleted by the user or until account disconnection.

Standard Retention Schedule: Scanned email forensic records remain accessible in your isolated dashboard workspace until you explicitly delete them or disconnect your account. Inactive account records and scan caches are retained for a maximum of 90 days.

D. User Data Deletion & Account Disconnection Rights

You have continuous, unilateral control over your stored records:

• Disconnecting Your Gmail Account: Clicking “Disconnect” in Settings or the Header menu immediately triggers two actions:
  1. The backend sends a token revocation request to Google's OAuth 2.0 revocation endpoint (https://oauth2.googleapis.com/revoke), terminating the authorization with Google; and
  2. Your account record and encrypted access/refresh tokens are permanently erased from our database.
• Deleting Individual Email Records: You can delete any individual scanned email from the Email History table at any time using the Delete action.
• Full Data Purge (“Reset Data”): You can instantly purge all scanned email records, metrics, and threat alerts from your workspace by using the Danger Zone > Reset Data action in Settings.
• Independent Revocation in Google: You may unilaterally revoke the application's access at any time via your Google Account Third-Party Access Portal.

6. Third-Party Sharing & Service Providers

We do not sell, rent, commercialize, or share your personal information. Data is communicated only with technical service providers essential to hosting and operating the platform:

• Google APIs (Google LLC): Facilitates OAuth 2.0 user authentication, session binding, and read-only message inspection via the official Gmail REST API.
• Vercel Inc.: Cloud edge hosting provider serving the client-side user interface under strict TLS/HTTPS.
• Render Services Inc.: Cloud platform provider hosting the backend FastAPI application and threat evaluation engine over encrypted HTTPS connections.
• MongoDB Inc. (MongoDB Atlas): Managed cloud database provider storing encrypted OAuth credentials and isolated user email scan history.

No External AI / LLM APIs or Threat Feeds: The Service operates its machine learning inference (Scikit-learn TF-IDF vectorization and Logistic Regression) and heuristic analysis entirely on-premises within our dedicated backend server. We do not transmit your emails, headers, or URLs to third-party AI APIs (such as OpenAI, Anthropic, or Google Gemini) or external security scanning services (such as VirusTotal).

We do not share data with any marketing providers, analytics brokers, or behavioral tracking networks.

7. Security Controls & Inherent Limitations

A. Technical Safeguards

• Encryption in Transit: All HTTP traffic, API endpoints, and real-time WebSocket streams (wss://) are enforced over SSL/TLS (HTTPS).
• Encryption at Rest: OAuth credentials are encrypted at rest with Fernet symmetric cryptography (AES-128-CBC / SHA-256 HMAC).
• Server-Side Identity Verification: Client-supplied user headers are never trusted. Every protected API endpoint verifies a server-signed JWT Bearer token.
• CORS Hardening: Production environments enforce strict Cross-Origin Resource Sharing (CORS) origin whitelisting.

B. Limitations & Security Disclaimer

While we implement industry-standard administrative and cryptographic safeguards, no method of transmission over the Internet or electronic storage is 100% secure. The Service operates as an informational security advisory tool; users remain responsible for maintaining general account security (e.g., multi-factor authentication on Google accounts).

8. Children's Privacy

The Service is not intended for or directed toward children under 13 years of age (or 16 years of age in the European Union). We do not knowingly collect personal data from children.

9. Modifications to this Privacy Policy

We may update this Privacy Policy from time to time to reflect operational, architectural, or legal updates. Material updates will be indicated by the “Last Updated” date at the top of this document. Continued use of the Service following revisions constitutes your acknowledgment of the updated policy.

10. Inquiries and Contact Information

For inquiries, data deletion requests, or questions regarding this Privacy Policy and our data handling practices, please contact:

• Project Repository: github.com/vibhu8791/AI-Email-Phising-Detector
• Issue Tracker: Submit a Privacy Request / Issue
• Contact Email: vibhuagarwal2006@gmail.com
• Developer / Entity Name: Vibhu Agarwal (Project Maintainer)
• Mailing Address / Jurisdiction: India