Cybersecurity

AI Zero-Trust Security System (AI-ZT-G)

AI-driven access control for a private company: no request is trusted until it is verified.

Every request scored for risk by AI

Sector
Cybersecurity
Client
Withheld · NDA
Built with
C# · ASP.NET Core · Avalonia · SQL

Problem

A private company holding sensitive data relied on access control that granted each user fixed permissions once, then trusted them for everything that followed:

  • Once a user had logged in, they were trusted everywhere, whatever the time, the device or the data they asked for.
  • A stolen password, or an employee acting out of pattern, went unnoticed.
  • Permissions were managed by hand, which led to human error.
  • There was no complete record of who accessed what, and when.
  • Common web attacks, such as SQL injection and cross-site scripting (XSS), remained a constant threat.

What we built

A security system built on the Zero-Trust principle: never trust, always verify. Every request is checked against its context and assessed by AI before access is granted.

  • AI at the core:
    • A risk score for every request. The AI assesses each request and gives it a risk score, then the system allows it, denies it or asks for extra verification.
    • Unusual-behaviour detection. The AI learns each user’s normal pattern and flags logins and activity that do not fit it.
  • Context-aware access. Before granting access, the system checks:
    • The user’s role. Permissions follow each user’s role.
    • The time. Access can be limited to set hours, such as working hours.
    • The device and network. The device and IP address the request comes from.
    • The sensitivity of the data. Requests for sensitive data require extra verification.
  • Control and visibility:
    • User management. Users are added, their roles changed and their access suspended from one place.
    • Session management. Sessions end automatically, and every active session can be seen and followed.
    • Activity log and monitoring. Every login and every action is recorded: who did what, and when.
    • Security alerts. Administrators are alerted the moment a suspicious login or request is detected.
    • Desktop app. Administrators manage the whole system from a cross-platform desktop application.
  • Security by design:
    • Encrypted passwords. Passwords are never stored in readable form.
    • Protection against attacks. Every input is validated to block SQL injection and XSS.
    • Three-tier architecture. Presentation, application and data are kept separate, so the system is easier to scale and maintain.
  • Tested before go-live. The system passed functional, performance and user acceptance testing (UAT) before it went into use.

Outcome

The company now manages access more efficiently, with fewer human errors and a stronger security posture. No request is trusted by default, unusual behaviour is caught as it happens, and every action is on record.

More work