Sample images can be independently restored from face recognition templates
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
Biometrics promise the ability to automatically identify individuals from reasonably easy to measure and hard to falsify characteristics. They are increasingly being investigated for use in large scale identification applications in the context of increased national security awareness. This paper addresses some of the security and privacy implications of biometric storage. Biometric systems record a sample image, and calculate a template: a compact digital representation of the essential features of the image. To compare the individuals represented by two images, the corresponding templates are compared, and a match score calculated, indicating the confidence level that the images represent the same individual. Biometrics vendors have uniformly claimed that it is impossible or infeasible to recreate an image from a template, and therefore, templates are currently treated as nonidentifiable data. We describe a simple algorithm which allows recreation of a sample image from a face recognition template using only match score values. At each iteration, a candidate image is slightly modified by an eigenface image, and modifications which improve the match score are kept. The regenerated image compares with high score to the original image, and visually shows most of the essential features. This image could thus be used to fool the algorithm as the target person, or to visually identify that individual. Importantly, this algorithm is immune to template encryption: any system which allows access to match scores effectively allows sample images to be regenerated in this way.
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 | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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