Marcus Rodriguez

General 11 min

Marcus Rodriguez • 1 week ago

How Code Plagiarism Detection Went From Hashes to LLMs

Ottenstein's 1976 detector hashed student Fortran token streams, and most of what we run today is a refined version of the same idea. This is the fifty-year arc from line diffs to winnowing, AST matching, web crawling, and statistical AI detection, plus the failure mode that still bites: a 0% similarity score that tells you nothing about authorship.

Read

General 11 min

Marcus Rodriguez • 1 week ago

A Framework for Verifying Code Originality From Contractors

Most statements of work say "original work" and never define it, which is how GPL code ends up in your settlement service. Here is the four-question intake review I run on every contractor deliverable, with the thresholds and tooling that hold up under scrutiny.

Read

General 9 min

Marcus Rodriguez • 4 weeks ago

How Code Plagiarism Detection Algorithms Ignore Renamed Variables

A student renames every variable and converts for loops to while loops. MOSS still flags 94%. This guide builds a minimal winnowing detector in Python 3.11 so you can see exactly why code plagiarism detection algorithms survive refactoring. We then look at where the approach breaks and how AST matching fills the gap.

Read

General 11 min

Marcus Rodriguez • 1 month ago

Web Code Plagiarism Detection Through Source Fingerprinting

Peer-based plagiarism checkers miss code copied from GitHub and Stack Overflow. This analysis walks through how web source fingerprinting works, what a 214-submission Java audit found, and where the approach breaks down. Includes a method comparison table and a practical review workflow.

Read

General 4 min

Marcus Rodriguez • 1 month ago

What 14,000 Python Submissions Reveal About AI Detection Errors

A three-semester case study at Briarwood University tracked 14,000 Python assignments through four AI code detectors. The result: false positive rates from 4% to 9% overall, spiking to 23% on common algorithmic patterns. This article breaks down the data, the code patterns that trigger false flags, and a practical workflow for balancing detection with fairness.

Read

General 8 min

Marcus Rodriguez • 1 month ago

Grading Assignments to Detect AI-Generated Code in Student Submissions

Most AI-generated code in student submissions goes unnoticed when instructors rely on intuition or a single detector. This research-style guide explains how to layer statistical signals, peer similarity, web-source checks, and rubric design to reliably catch AI-assisted code without manufacturing false positives.

Read

General 1 min

Marcus Rodriguez • 1 month ago

How Greedy String Tiling Detects Plagiarized Code

Greedy string tiling is the matching algorithm behind JPlag and several modern code similarity engines. This report explains how it tokenizes source, extracts maximal contiguous matches, and why it still needs help from AST and fingerprinting to catch refactored plagiarism.

Read

General 11 min

Marcus Rodriguez • 2 months ago

At What Point Does Token-Based Detection Fail Against Automated Refactoring?

Most plagiarism detectors rely on token streams, which break down as soon as students rename variables, reorder statements, or extract methods. We map the precise failure points, walk through AST-based recovery techniques, and show how fingerprinting fills the gaps that tree comparators leave behind. A must-bookmark deep‑dive for any CS educator or engineering lead who has watched suspect code sail right through a token‑only scanner.

Read

General 10 min

Marcus Rodriguez • 2 months ago

15,000 CS Submissions Test 3 Plagiarism Detection Algorithms

Code similarity tools all promise to catch cheaters, but their underlying algorithms differ dramatically. We ran a large-scale experiment—15,000 real CS1 Java submissions, 500 manually verified suspicious pairs—to compare winnowing (MOSS), AST hashing (JPlag-style), and fingerprinting side by side. The results expose which techniques survive renaming, refactoring, and template reuse, and why a layered approach matters for low false‑positive rates in production academic workflows.

Read

General 10 min

Marcus Rodriguez • 2 months ago

How Perplexity and Burstiness Reveal AI-Written Code

AI code detectors don't read code—they measure its statistical shape. This piece breaks down the two key metrics, perplexity and burstiness, that separate lines from a language model from something a programmer actually typed. Real numbers, real edge cases, and how to combine signals for a higher-confidence verdict.

Read

General 12 min

Marcus Rodriguez • 2 months ago

Automating Source Code Plagiarism Checks With Canvas and Codequiry

When a single CS1 assignment yields 300+ submissions, manual plagiarism checking simply doesn't scale. This hands-on guide walks through connecting Canvas to Codequiry's API, running similarity and AI-detection scans with a handful of Python scripts, and posting flagged results directly back into the SpeedGrader — so you catch the cases that matter without drowning in paperwork.

Read

General 12 min

Marcus Rodriguez • 2 months ago

A Triage Protocol for AI-Generated Code in CS Assignments

A single run of an AI detector on a suspicious student submission is not enough. CS professors need a systematic triage protocol that stacks similarity analysis, AI code detection, web-source fingerprinting, and manual review into a defensible pipeline. This article outlines a concrete workflow you can implement this semester.

Read