SafeDocs

Intelligent Document-Based Malware Protection

Client

SafeDocs Security

Role

Full Stack Architecture & AI

Year

2026

Live

Visit Site
Project Hero Showcase

The Challenge

SafeDocs is an intelligent cybersecurity platform designed to protect users from document-based malware. It allows users to upload files which are instantly analyzed by a hybrid AI engine. The system determines a "Verdict" (Malicious or Benign) based on concrete evidence—such as the presence of obfuscated JavaScript or VBA macros—rather than just arbitrary risk scores. Malicious files are automatically "sanitized" (cleaned) and returned to the user safe to open.

01

The Problem

Traditional antivirus tools often rely on arbitrary "risk scores" based on file complexity, leading to high false positives for legitimate business documents. Users are often told a file is "suspicious" without explanation and have no way to recover the safe content from an infected file without destroying it.

02

The Solution

We engineered a hybrid analysis engine combining LightGBM, Deep Learning, and signature matching. SafeDocs moves beyond simple detection by implementing "Smart Sanitization"—surgically removing only the malicious elements (like hidden macros or scripts) to reconstruct a safe, usable version of the original file.

03

The Impact

he platform solves the "False Positive" dilemma by prioritizing concrete evidence over generic complexity scores. It delivers enterprise-grade security that doesn't just block work, but actively recovers safely usable documents from potentially infected sources.

Key Features

Multi-Model Analysis

Utilizes a voting system with LightGBM, Deep Learning, and Rule Engine (6000+ signatures).

Smart Sanitization

Automatically strips malicious elements (macros, JS) while preserving safe document integrity.

Evidence-Based Verdicts

Logic-driven decisions prioritizing high-severity findings over generic risk scores.

Real-time Dashboard

Responsive UI for tracking scan history, viewing risk scores, and downloading cleaned files.

Secure Infrastructure

Isolated processing environment with automatic 48-hour data purging for privacy.

False Positive Reduction

Advanced logic that prevents complex but benign files (like invoices) from being flagged incorrectly.

Project Highlight

Technologies Used

Next.js 16TypeScriptTailwind CSS & Framer MotionPython (FastAPI)PostgreSQL & PrismaLightGBM & Deep LearningArcjet SecurityRailway & Vercel
Project Finale