01 / Key message
Fighting Fire with Fire: On the Feasibility of Protecting Exercises Against AI Cheating
Protected visual questions steer AI assistants toward controlled wrong answers that form a detectable assignment-level fingerprint.
Assignment fingerprint
02 / Method
How it works
Choose a target
Assign each multimodal question a secret, incorrect target answer.
Steer the solver
Optimize a subtle visual perturbation against an ensemble of accessible surrogate models.
Calibrate transfer
Query frontier assistants repeatedly and retain question-model pairs with reliable target separation.
Detect the pattern
Combine retained questions into an assignment and test for unusual overlap with the hidden targets.
03 / Abstract
Abstract
As multimodal assistants solve more educational exercises, detecting copied answers after submission becomes increasingly unreliable. This work explores a preventive alternative: add subtle, task-preserving perturbations to the visual parts of multiple-choice questions so AI solvers are steered toward designated incorrect answers. Across an assignment, those controlled errors form a statistical fingerprint. The method optimizes against accessible surrogate models, calibrates transfer through repeated black-box queries, and assembles questions whose answer patterns support likelihood-ratio testing across Claude, Gemini, and GPT assistants. The study establishes feasibility under a defined sustained-copying threat model while making educator judgment and the method’s limitations explicit.
04 / Contributions
What this adds
- 01
Assessment-side intervention
Moves the problem from post-hoc authorship classification to proactive exercise design.
- 02
Controlled error fingerprint
Uses target-specific visual perturbations to create a pattern that blind AI copying reproduces across an assignment.
- 03
Calibrated detection
Combines black-box response calibration with assistant-specific statistical tests and clearly stated student-model assumptions.
05 / Citation
Citation
Braun, T., Grebe, J., Rethfeld, L., & Rohrbach, M. (2026). Fighting Fire with Fire: On the Feasibility of Protecting Exercises Against AI Cheating. Preprint.
BibTeX
@misc{braun2026fighting,
title = {Fighting Fire with Fire: On the Feasibility of Protecting Exercises Against AI Cheating},
author = {Tobias Braun and Jonas Grebe and Louis Rethfeld and Marcus Rohrbach},
year = {2026},
note = {Preprint}
}