SECURING...
Academic FBI provides a complete pipeline from student submission to similarity analysis — with encrypted storage, anonymous review, and cross-year comparison. No external AI services. No code execution.
AST-based code comparison using tree-sitter tokenization, adaptive k-gram hashing, and IDF-weighted Jaccard similarity. Detects structural plagiarism even when variable names, comments, and formatting are changed.
Review flagged pairs in a responsive split-panel viewer with block highlighting, confidence-level coloring (HIGH / MEDIUM / LOW), hover tooltips, and block-by-block navigation.
Upload previous years' submissions as reference repositories. The engine compares current students against historical work to catch recycled assignments across semesters.
Upload instructor-provided starter code as boilerplate. The engine fingerprints template code and excludes matching patterns from similarity scores, eliminating false positives.
All analysis results use anonymous labels (Student A, Student B). Real student identities are encrypted and only disclosed through a formal reveal request approved by the university admin.
Student files are AES-128 encrypted at rest using Fernet. Student identity fields are encrypted in the database. Submission folders use hashed IDs instead of plaintext student numbers.
Students upload code as ZIP files through a token-authenticated portal. No account required — just an assignment key from the instructor. Supports C, C++, and Java.
At-a-glance overview of courses, submissions, and flagged pairs with severity breakdowns. Quick-launch analysis directly from the dashboard without navigating to individual assignments.
Courses have configurable end dates with a 30-day grace period. After expiry, submission data is automatically purged to comply with data retention policies and free storage.