Investigating the Impact of Inclusion in Face Recognition Training Data\n on Individual Face Identification
Bibliographic record
Abstract
Modern face recognition systems leverage datasets containing images of\nhundreds of thousands of specific individuals' faces to train deep\nconvolutional neural networks to learn an embedding space that maps an\narbitrary individual's face to a vector representation of their identity. The\nperformance of a face recognition system in face verification (1:1) and face\nidentification (1:N) tasks is directly related to the ability of an embedding\nspace to discriminate between identities. Recently, there has been significant\npublic scrutiny into the source and privacy implications of large-scale face\nrecognition training datasets such as MS-Celeb-1M and MegaFace, as many people\nare uncomfortable with their face being used to train dual-use technologies\nthat can enable mass surveillance. However, the impact of an individual's\ninclusion in training data on a derived system's ability to recognize them has\nnot previously been studied. In this work, we audit ArcFace, a\nstate-of-the-art, open source face recognition system, in a large-scale face\nidentification experiment with more than one million distractor images. We find\na Rank-1 face identification accuracy of 79.71% for individuals present in the\nmodel's training data and an accuracy of 75.73% for those not present. This\nmodest difference in accuracy demonstrates that face recognition systems using\ndeep learning work better for individuals they are trained on, which has\nserious privacy implications when one considers all major open source face\nrecognition training datasets do not obtain informed consent from individuals\nduring their collection.\n
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.
How this classification was reachedexpand
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 | 0.002 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.004 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.005 | 0.013 |
| Research integrity | 0.000 | 0.001 |
| 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".