Skip to content

Add face recognition benchmark pages #14

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
vboddeti wants to merge 9 commits into
base: main
Choose a base branch
from
Draft

Add face recognition benchmark pages #14

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
10635ef
Add face recognition benchmark pages
vboddeti Aug 5, 2026
16ed0c6
Refresh face recognition results
vboddeti Aug 6, 2026
a62122b
Add FHE learning resources
vboddeti Aug 6, 2026
494cb75
Add publication details to resources
vboddeti Aug 6, 2026
9fa40d2
Deduplicate resource links
vboddeti Aug 6, 2026
784fd9e
Order resources by publication year
vboddeti Aug 6, 2026
226fc1b
Fix local preview and result headers
vboddeti Aug 6, 2026
caeb7ea
Add resources to face recognition page
vboddeti Aug 6, 2026
36cc20d
Polish face recognition website pages
vboddeti Aug 8, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 13 additions & 1 deletion README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,6 +32,8 @@ Each workload represents a real-world use case with clear specifications, refere

- [**BERT**](./BERT/index.html). Privacy-preserving BERT inference on encrypted inputs. Currently supports [BERT-Base (110M)](https://huggingface.co/google-bert/bert-base-cased-finetuned-mrpc) inference on the MRPC task in the [GLUE benchmark](https://gluebenchmark.com/).

- [**Face Recognition**](./face-recognition/index.html). End-to-end privacy-preserving face verification with encrypted feature extraction and matching, including single-pair and batched benchmark results.

*Additional workloads are under development. Check back for updates.*

## Suite Structure
Expand Down Expand Up @@ -84,6 +86,16 @@ For questions or to get involved:
Visit the Homomorphic Encryption Standardization community:
[homomorphicencryption.org](https://homomorphicencryption.org)

Additional tutorials and research:

- [Computer Vision over Homomorphically Encrypted Data — CVPR 2025 Tutorial](https://fhe4cv.github.io/)
- [SecureRAG: End-to-End Secure Retrieval-Augmented Generation](https://openreview.net/pdf?id=5uXACIHz6K) — *NeurIPS GenAI4Health Workshop, 2025*.
- [Homomorphically Encrypted Biometric Template Fusion and Matching](https://doi.org/10.1109/TBIOM.2025.3595438) — *IEEE Transactions on Biometrics, Behavior, and Identity Science, 2025*.
- [AutoFHE: Automated Adaptation of CNNs for Efficient Evaluation over FHE](https://arxiv.org/abs/2310.08012) — *33rd USENIX Security Symposium, 2024*.
- [HERS: Homomorphically Encrypted Representation Search](https://arxiv.org/abs/2003.12197) — *IEEE Transactions on Biometrics, Behavior, and Identity Science, 2022*.
- [HEFT: Homomorphically Encrypted Fusion of Biometric Templates](https://arxiv.org/abs/2208.07241) — *IEEE International Joint Conference on Biometrics, 2022*.
- [Secure Face Matching Using Fully Homomorphic Encryption](https://arxiv.org/abs/1805.00577) — *IEEE International Conference on Biometrics: Theory, Applications, and Systems, 2018*.

### Organizers
Andreea Alexandru, Flavio Bergamaschi, Shruthi Gorantala, Shai Halevi

Expand All @@ -98,4 +110,4 @@ Andreea Alexandru, Flavio Bergamaschi, Shruthi Gorantala, Shai Halevi

---

[Code of Conduct](./CODE_OF_CONDUCT.html) | © 2026 [HomomorphicEncryption.org](https://homomorphicencryption.org) | Last updated 2026-02-19
[Code of Conduct](./CODE_OF_CONDUCT.html) | © 2026 [HomomorphicEncryption.org](https://homomorphicencryption.org) | Last updated 2026-08-06
169 changes: 169 additions & 0 deletions face-recognition/Large.html
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,169 @@
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<link rel="icon" href="data:,">
<title>FHE Benchmarking Results - Face Recognition (Large Batch - 1,024)</title>
<link rel="stylesheet" href="https://cdn.datatables.net/1.13.6/css/jquery.dataTables.min.css">
<link rel="stylesheet" href="https://cdn.datatables.net/buttons/2.4.2/css/buttons.dataTables.min.css">
<script src="https://code.jquery.com/jquery-3.7.0.min.js"></script>
<script src="https://cdn.datatables.net/1.13.6/js/jquery.dataTables.min.js"></script>
<script src="https://cdn.datatables.net/buttons/2.4.2/js/dataTables.buttons.min.js"></script>
<script src="https://cdn.datatables.net/buttons/2.4.2/js/buttons.colVis.min.js"></script>
<style>
body { font-family: Arial, sans-serif; margin: 0; padding: 16px; color: #17202a; }
h1 { font-size: 1.5rem; letter-spacing: 0; }
.table-wrap { width: 100%; overflow-x: auto; }
#resultsTable { border-collapse: collapse; width: 100%; }
.table-wrap table.dataTable th, .table-wrap table.dataTable td { padding: 4px 8px; white-space: nowrap; }
.table-wrap table.dataTable thead tr:first-child th { color: white; text-align: center; }
.table-wrap table.dataTable thead tr:nth-child(2) th { color: white; font-size: 0.85rem; }
.group-header { text-align: center !important; vertical-align: middle; }
.submitter-header { background: #2471a3 !important; }
.bandwidth-header { background: #1e8449 !important; }
.quality-header { background: #7d3c98 !important; }
.harness-timing-header { background: #b05d0b !important; }
.server-timing-header { background: #2874a6 !important; }
#resultsTable tbody tr:nth-child(odd) { background: #edf2f4; }
.dt-buttons { margin-bottom: 10px; }
</style>
</head>
<body>
<h1>FHE Benchmarking Results - Face Recognition (Large Batch - 1,024)</h1>
<p>Timing and batched quality values are averages across 3 measurement run(s). Single-pair scores and labels are listed per run. Durations are wall time unless labeled as worker time.</p>
<div class="table-wrap">
<table id="resultsTable" class="display">
<thead>
<tr>
<th class="submitter-header group-header" colspan="9">Submitter</th>
<th class="bandwidth-header group-header" colspan="4">Bandwidth</th>
<th class="quality-header group-header" colspan="11">Quality</th>
<th class="harness-timing-header group-header" colspan="14">Timing (harness)</th>
<th class="server-timing-header group-header" colspan="13">Timing (server)</th>
</tr>
<tr>
<th class="submitter-header">Name</th>
<th class="submitter-header">Date</th>
<th class="submitter-header">CPU</th>
<th class="submitter-header">RAM</th>
<th class="submitter-header">Orion</th>
<th class="submitter-header">Slots</th>
<th class="submitter-header">Chunk</th>
<th class="submitter-header">R/L</th>
<th class="submitter-header">Runs</th>
<th class="bandwidth-header">Keys</th>
<th class="bandwidth-header">Model</th>
<th class="bandwidth-header">Input</th>
<th class="bandwidth-header">Result</th>
<th class="quality-header">FHE EER</th>
<th class="quality-header">FHE TAR@1%</th>
<th class="quality-header">FHE TAR@0.1%</th>
<th class="quality-header">ArcFace EER</th>
<th class="quality-header">ArcFace TAR@1%</th>
<th class="quality-header">ArcFace TAR@0.1%</th>
<th class="quality-header">EER gap</th>
<th class="quality-header">TAR@1% gap</th>
<th class="quality-header">TAR@0.1% gap</th>
<th class="quality-header">Max EER gap</th>
<th class="quality-header">Result</th>
<th class="harness-timing-header">Total</th>
<th class="harness-timing-header">Offline setup</th>
<th class="harness-timing-header">Online evaluation</th>
<th class="harness-timing-header">Dataset</th>
<th class="harness-timing-header">Keygen</th>
<th class="harness-timing-header">Model prep</th>
<th class="harness-timing-header">Input gen</th>
<th class="harness-timing-header">Input prep</th>
<th class="harness-timing-header">Input enc</th>
<th class="harness-timing-header">Compute</th>
<th class="harness-timing-header">Decrypt</th>
<th class="harness-timing-header">Postprocess</th>
<th class="harness-timing-header">ArcFace</th>
<th class="harness-timing-header">Quality</th>
<th class="server-timing-header">Total (wall)</th>
<th class="server-timing-header">Process lifetime</th>
<th class="server-timing-header">Setup (wall)</th>
<th class="server-timing-header">Model I/O</th>
<th class="server-timing-header">Runtime compile</th>
<th class="server-timing-header">Input I/O (worker)</th>
<th class="server-timing-header">Transport (worker)</th>
<th class="server-timing-header">Inference (worker)</th>
<th class="server-timing-header">Compute (wall)</th>
<th class="server-timing-header">Backbones (worker)</th>
<th class="server-timing-header">Normalize (worker)</th>
<th class="server-timing-header">Inner product (worker)</th>
<th class="server-timing-header">Mean / pair</th>
</tr>
</thead>
<tbody>
<tr>
<td><a href="https://github.com/fhe-benchmarking/face-recognition">CryptoFace</a></td>
<td>2026-08-06 11:57:25</td>
<td>AMD EPYC 7502 32-Core Processor</td>
<td>1007.7 GiB</td>
<td>7e27281354f9</td>
<td>10</td>
<td>50</td>
<td>L</td>
<td>3</td>
<td>28.8G</td>
<td>114.1G</td>
<td>128.1G</td>
<td>1.0G</td>
<td>0.0723</td>
<td>0.8477</td>
<td>0.8203</td>
<td>0.0605</td>
<td>0.9355</td>
<td>0.9316</td>
<td>0.0117</td>
<td>-0.0879</td>
<td>-0.1113</td>
<td>0.1500</td>
<td>PASS</td>
<td>12.507h</td>
<td>17.689m</td>
<td>12.213h</td>
<td>325.8ms</td>
<td>8.320m</td>
<td>9.364m</td>
<td>19.437s</td>
<td>22.180m</td>
<td>46.576m</td>
<td>10.688h</td>
<td>3.428m</td>
<td>47.4ms</td>
<td>18.950m</td>
<td>11.7ms</td>
<td>10.644h</td>
<td>10.677h</td>
<td>22.475m</td>
<td>13.842m</td>
<td>8.478m</td>
<td>51.956m</td>
<td>9.143m</td>
<td>663.424h</td>
<td>10.270h</td>
<td>662.592h</td>
<td>42.366m</td>
<td>7.560m</td>
<td>36.105s</td>
</tr>
</tbody>
</table>
</div>
<script>
$(document).ready(function() {
$('#resultsTable').DataTable({
dom: 'Bfrtip',
buttons: [{ extend: 'colvis' }],
paging: false,
searching: false,
info: false,
scrollX: true
});
});
</script>
</body>
</html>
169 changes: 169 additions & 0 deletions face-recognition/Medium.html
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,169 @@
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<link rel="icon" href="data:,">
<title>FHE Benchmarking Results - Face Recognition (Medium Batch - 256)</title>
<link rel="stylesheet" href="https://cdn.datatables.net/1.13.6/css/jquery.dataTables.min.css">
<link rel="stylesheet" href="https://cdn.datatables.net/buttons/2.4.2/css/buttons.dataTables.min.css">
<script src="https://code.jquery.com/jquery-3.7.0.min.js"></script>
<script src="https://cdn.datatables.net/1.13.6/js/jquery.dataTables.min.js"></script>
<script src="https://cdn.datatables.net/buttons/2.4.2/js/dataTables.buttons.min.js"></script>
<script src="https://cdn.datatables.net/buttons/2.4.2/js/buttons.colVis.min.js"></script>
<style>
body { font-family: Arial, sans-serif; margin: 0; padding: 16px; color: #17202a; }
h1 { font-size: 1.5rem; letter-spacing: 0; }
.table-wrap { width: 100%; overflow-x: auto; }
#resultsTable { border-collapse: collapse; width: 100%; }
.table-wrap table.dataTable th, .table-wrap table.dataTable td { padding: 4px 8px; white-space: nowrap; }
.table-wrap table.dataTable thead tr:first-child th { color: white; text-align: center; }
.table-wrap table.dataTable thead tr:nth-child(2) th { color: white; font-size: 0.85rem; }
.group-header { text-align: center !important; vertical-align: middle; }
.submitter-header { background: #2471a3 !important; }
.bandwidth-header { background: #1e8449 !important; }
.quality-header { background: #7d3c98 !important; }
.harness-timing-header { background: #b05d0b !important; }
.server-timing-header { background: #2874a6 !important; }
#resultsTable tbody tr:nth-child(odd) { background: #edf2f4; }
.dt-buttons { margin-bottom: 10px; }
</style>
</head>
<body>
<h1>FHE Benchmarking Results - Face Recognition (Medium Batch - 256)</h1>
<p>Timing and batched quality values are averages across 3 measurement run(s). Single-pair scores and labels are listed per run. Durations are wall time unless labeled as worker time.</p>
<div class="table-wrap">
<table id="resultsTable" class="display">
<thead>
<tr>
<th class="submitter-header group-header" colspan="9">Submitter</th>
<th class="bandwidth-header group-header" colspan="4">Bandwidth</th>
<th class="quality-header group-header" colspan="11">Quality</th>
<th class="harness-timing-header group-header" colspan="14">Timing (harness)</th>
<th class="server-timing-header group-header" colspan="13">Timing (server)</th>
</tr>
<tr>
<th class="submitter-header">Name</th>
<th class="submitter-header">Date</th>
<th class="submitter-header">CPU</th>
<th class="submitter-header">RAM</th>
<th class="submitter-header">Orion</th>
<th class="submitter-header">Slots</th>
<th class="submitter-header">Chunk</th>
<th class="submitter-header">R/L</th>
<th class="submitter-header">Runs</th>
<th class="bandwidth-header">Keys</th>
<th class="bandwidth-header">Model</th>
<th class="bandwidth-header">Input</th>
<th class="bandwidth-header">Result</th>
<th class="quality-header">FHE EER</th>
<th class="quality-header">FHE TAR@1%</th>
<th class="quality-header">FHE TAR@0.1%</th>
<th class="quality-header">ArcFace EER</th>
<th class="quality-header">ArcFace TAR@1%</th>
<th class="quality-header">ArcFace TAR@0.1%</th>
<th class="quality-header">EER gap</th>
<th class="quality-header">TAR@1% gap</th>
<th class="quality-header">TAR@0.1% gap</th>
<th class="quality-header">Max EER gap</th>
<th class="quality-header">Result</th>
<th class="harness-timing-header">Total</th>
<th class="harness-timing-header">Offline setup</th>
<th class="harness-timing-header">Online evaluation</th>
<th class="harness-timing-header">Dataset</th>
<th class="harness-timing-header">Keygen</th>
<th class="harness-timing-header">Model prep</th>
<th class="harness-timing-header">Input gen</th>
<th class="harness-timing-header">Input prep</th>
<th class="harness-timing-header">Input enc</th>
<th class="harness-timing-header">Compute</th>
<th class="harness-timing-header">Decrypt</th>
<th class="harness-timing-header">Postprocess</th>
<th class="harness-timing-header">ArcFace</th>
<th class="harness-timing-header">Quality</th>
<th class="server-timing-header">Total (wall)</th>
<th class="server-timing-header">Process lifetime</th>
<th class="server-timing-header">Setup (wall)</th>
<th class="server-timing-header">Model I/O</th>
<th class="server-timing-header">Runtime compile</th>
<th class="server-timing-header">Input I/O (worker)</th>
<th class="server-timing-header">Transport (worker)</th>
<th class="server-timing-header">Inference (worker)</th>
<th class="server-timing-header">Compute (wall)</th>
<th class="server-timing-header">Backbones (worker)</th>
<th class="server-timing-header">Normalize (worker)</th>
<th class="server-timing-header">Inner product (worker)</th>
<th class="server-timing-header">Mean / pair</th>
</tr>
</thead>
<tbody>
<tr>
<td><a href="https://github.com/fhe-benchmarking/face-recognition">CryptoFace</a></td>
<td>2026-08-05 19:34:36</td>
<td>AMD EPYC 7502 32-Core Processor</td>
<td>1007.7 GiB</td>
<td>7e27281354f9</td>
<td>10</td>
<td>50</td>
<td>L</td>
<td>3</td>
<td>28.8G</td>
<td>114.1G</td>
<td>32.0G</td>
<td>258.6M</td>
<td>0.0868</td>
<td>0.8395</td>
<td>0.8208</td>
<td>0.0721</td>
<td>0.9256</td>
<td>0.9209</td>
<td>0.0146</td>
<td>-0.0861</td>
<td>-0.1001</td>
<td>0.1500</td>
<td>PASS</td>
<td>3.418h</td>
<td>21.038m</td>
<td>3.067h</td>
<td>415.5ms</td>
<td>8.578m</td>
<td>12.453m</td>
<td>5.577s</td>
<td>4.611m</td>
<td>12.660m</td>
<td>2.670h</td>
<td>1.675m</td>
<td>25.4ms</td>
<td>4.779m</td>
<td>64.8ms</td>
<td>2.633h</td>
<td>2.664h</td>
<td>20.390m</td>
<td>12.060m</td>
<td>8.221m</td>
<td>8.295m</td>
<td>1.903m</td>
<td>143.533h</td>
<td>2.294h</td>
<td>143.339h</td>
<td>9.848m</td>
<td>1.763m</td>
<td>32.254s</td>
</tr>
</tbody>
</table>
</div>
<script>
$(document).ready(function() {
$('#resultsTable').DataTable({
dom: 'Bfrtip',
buttons: [{ extend: 'colvis' }],
paging: false,
searching: false,
info: false,
scrollX: true
});
});
</script>
</body>
</html>
Loading