ExamGuard-X

Inspiration

Question-paper leaks in examinations remain a critical vulnerability, disrupting educational systems and affecting millions of students. Traditional security measures often rely on isolated barriers—such as password-protected databases, physical locks, or post-exam investigations—without a unified mechanism to secure the entire lifecycle of a paper from creation to distribution and detection. Inspired by the need for robust, multi-layered transparency and accountability, we built ExamGuard-X.

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for ExamGuard-X

ExamGuard-X is an AI-powered zero-trust platform designed to protect, detect, trace, and investigate question-paper leaks across both online and offline examination environments.Secure Lifecycle Management: Handles creation, approval, encryption, and scheduled release of question papers.Zero-Trust Access Control: Evaluates requests using identity, roles, devices, and security risk factors rather than relying solely on perimeter security.AI Leak Detection: Uses image preprocessing, OCR (PaddleOCR), and semantic embeddings (Sentence Transformers + FAISS) to scan leaked PDFs or photos for complete, partial, or rephrased matches.Forensic Traceability: Implements unique watermarking, print records, and an auditable chain-of-custody tracking framework.Investigation Dashboard: Connects users, papers, devices, and events through an investigation graph (Neo4j) to assist authorized personnel.How we built itFrontend: Built with React, TypeScript, and Tailwind CSS for a responsive and intuitive user interface. Backend & Security: Powered by Java Spring Boot and Spring Security handling REST APIs, business logic, and zero-trust controls. Databases & Streaming: Utilizes MySQL for structured relational data, Redis for caching/temporary data, and Apache Kafka for event-driven security streaming. AI & Analytics Services: Developed using Python with FastAPI, integrating PaddleOCR, OpenCV, Sentence Transformers, FAISS, and Scikit-learn for anomaly and similarity detection. Graph & Infrastructure: Uses Neo4j for tracking relationship chains during investigations, containerized via Docker and deployed on AWS. Challenges we ran intoSynchronizing real-time event streaming across Kafka while maintaining strict access-control guarantees.Optimizing semantic similarity searches using FAISS to quickly match partial, rephrased, or OCR-distorted text against original question banks.Designing a comprehensive chain-of-custody tracking model that accurately reflects physical printing and movement stages without slowing down system throughput.What we learnedThe importance of a defense-in-depth architecture where cryptographic protection, behavioral anomaly detection, and zero-trust principles work harmoniously.How combining structured relational data with graph databases (Neo4j) radically simplifies complex forensic investigations.What's next for ExamGuard-XConducting extensive real-world performance evaluations and benchmarking using larger synthetic datasets.Refining automated alert thresholds to further minimize false-positive rates during behavioral anomaly checks.Feel free to paste this into your Devpost Project Story box! Let me know if you need any tweaks or help filling out the remaining sections.

Built With

• ai
• java
• springboot