A Cloud-based Ai-powered Threat Deception Platform

Abstract

Modern Web Applications Are Increasingly Targeted By Automated Bots And Sophisticated Attackers Using Advanced Exploitation Techniques Such As Injection Attacks, Credential Stuffing, And Reconnaissance-based Probing. Traditional Intrusion Detection Systems Primarily Focus On Detection And Blocking, Often Failing To Extract Actionable Intelligence From Adversarial Interactions. This Paper Presents A Cloud-based, AI-powered Threat Deception Platform That Actively Engages Attackers Through Realistic Honeypot Interfaces And Tarpit Mechanisms While Simultaneously Analyzing Behavioral And Payload-level Data. The Proposed System Integrates Rule-based Attack Signature Detection With An XGBoost-based Behavioral Machine Learning Model To Identify Malicious Activity With High Accuracy. Severity Assessment Is Performed Using CVSS 3.1 Scoring, And Detected Threats Are Mapped To OWASP Top 10 Categories And Relevant CVE References. The Platform Is Fully Deployed On Cloud Infrastructure Using Firebase Hosting, A Flask-based Backend, And Azure Blob Storage For Scalable Logging. Experimental Evaluation Demonstrates Effective Detection Of Multiple Attack Vectors Including XSS, SQL Injection, Command Injection, And Automated Bot Behavior, While Maintaining Low Operational Cost. The Results Indicate That The Proposed System Not Only Detects Threats But Also Converts Attacks Into Valuable Security Intelligence.

Keywords