A Psycholinguistics-inspired Method to Counter IP Theft Using Fake Documents
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Intellectual property (IP) theft is a growing problem. We build on prior work to deter IP theft by generating n fake versions of a technical document so a thief has to expend time and effort in identifying the correct document. Our new SbFAKE framework proposes, for the first time, a novel combination of language processing, optimization, and the psycholinguistic concept of surprisal to generate a set of such fakes. We start by combining psycholinguistic-based surprisal scores and optimization to generate two bilevel surprisal optimization problems (an Explicit one and a simpler Implicit one) whose solutions correspond directly to the desired set of fakes. As bilevel problems are usually hard to solve, we then show that these two bilevel surprisal optimization problems can each be reduced to equivalent surprisal-based linear programs. We performed detailed parameter tuning experiments and identified the best parameters for each of these algorithms. We then tested these two variants of SbFAKE (with their best parameter settings) against the best performing prior work in the field. Our experiments show that SbFAKE is able to more effectively generate convincing fakes than past work. In addition, we show that replacing words in an original document with words having similar surprisal scores generates greater levels of deception.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it