Skip to content

Commit 258595d

Browse files
committed
update: refine social and GitHub links in README
1 parent 14cbf3a commit 258595d

File tree

6 files changed

+40
-3
lines changed

6 files changed

+40
-3
lines changed

.gitignore

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,2 @@
1+
.DS_Store
2+
__pycache__/

README.md

Lines changed: 38 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,38 @@
1+
<div align="center">
2+
3+
<!-- Logo with light mode support -->
4+
<picture>
5+
<source media="(prefers-color-scheme: light)" srcset="docs/nesa-transparent-logo.png">
6+
<img alt="Nesa Logo" src="docs/nesa-logo.png" width="50%">
7+
</picture>
8+
<br>
9+
<br>
10+
<p>
11+
<em><code>Witness the power of trusted AI on chain.</code></em>
12+
</p>
13+
14+
<h3 style="margin-top: 15px; margin-bottom: 15px; display: flex; justify-content: center; align-items: center; gap: 15px;">
15+
<a href="https://discord.gg/TK89MgJDkz" style="text-decoration: none;">
16+
<img alt="Discord" src="https://img.shields.io/badge/-Discord-7289DA?style=flat&logo=discord&logoColor=white"></a>
17+
<a href="https://x.com/nesaorg/" style="text-decoration: none;">
18+
<img alt="X" src="https://img.shields.io/twitter/follow/nesaorg?style=social">
19+
</a>
20+
</h3>
21+
<!-- GitHub Repo Stats -->
22+
<div>
23+
<a href="https://github.com/nesaorg/Equivariant-Encryption-for-AI/stargazers" target="_blank">
24+
<img alt="GitHub Stars" src="https://img.shields.io/static/v1?label=Stars&message=201&color=brightgreen&logo=github&style=social"></a>
25+
<a href="https://github.com/nesaorg" style="text-decoration: none;">
26+
<img alt="GitHub Follow" src="https://img.shields.io/github/followers/nesaorg?label=follow&style=social">
27+
</a>
28+
</div>
29+
</div>
30+
31+
<!---
32+
---
33+
-->
34+
35+
136
# Equivariant Encryption for AI
237

338
At Nesa, privacy is a critical objective. On our path toward decentralized AI, we confronted a key challenge: **how can we perform inference on neural networks without exposing the underlying input and output data to external parties?** Traditional approaches, such as differential privacy or homomorphic encryption (HE), while conceptually strong, fall short in practical deployments for complex neural architectures. These methods can struggle to handle non-linear operations efficiently, often imposing substantial computational overhead that makes them challenging to integrate into real-time or large-scale systems.
@@ -30,7 +65,7 @@ Despite this, the conceptual promise of HE—running inference on encrypted data
3065
Our solution is **Equivariant Encryption (EE)**. The term **equivariance** signifies a change in representation that preserves the operational structure from the model’s perspective. In other words, we transform the input data into an encrypted domain where the neural network’s computations can be carried out as though it were processing plaintext, all while maintaining the secrecy of the underlying information.
3166

3267
<div align="center">
33-
<img src="ee.png" alt="equivariant encryption diagram">
68+
<img src="docs/ee.png" alt="equivariant encryption diagram">
3469
</div>
3570

3671
Rather than relying exclusively on arithmetic operations compatible with HE, EE integrates specialized transformations designed around the internal properties of neural networks. We exploit the known architecture, layer composition, and input-output mappings of the model to construct a system in which each step of inference operates correctly on encrypted inputs. This approach avoids expensive retraining on encrypted datasets. Instead, by following a set of mathematical guidelines, we can generate a new variant of the model that works with our encryption schema in a matter of seconds.
@@ -42,7 +77,7 @@ Crucially, the complexity of inference under EE does not surpass that of the une
4277
To illustrate this with a tangible example, consider transformer-based models like ChatGPT, Claude, or Llama. These models employ tokenizers to convert text into discrete tokens, each mapped to an integer token ID. Under EE, we implement a specialized tokenizer that produces a different, encrypted set of token IDs. The network, now adapted to EE, treats these encrypted token IDs as standard inputs. It processes them identically to how it would process normal tokens, ultimately returning encrypted output tokens that can be decrypted locally by the user. The following diagram outlines this workflow:
4378

4479
<div align="center">
45-
<img src="tokenizer.png" alt="tokenizer diagram">
80+
<img src="docs/tokenizer.png" alt="tokenizer diagram">
4681
</div>
4782

4883
In this setup, all data traveling over the network remains encrypted, and the transformations that produce and consume these tokens are carefully chosen to deny any straightforward method for recovering the plaintext. The attacker sees only encrypted tokens and a model variant designed to operate on that encrypted space, providing no direct, low-cost avenue to extract the original information.
@@ -69,4 +104,4 @@ Below is a more detailed breakdown of how Equivariant Encryption matches or outp
69104
We invite the community to examine and test the security claims of our Equivariant Encryption approach. As part of our commitment to transparency and continual refinement, we have organized a contest encouraging participants to probe for weaknesses and demonstrate potential exploits.
70105

71106
For details, please visit:
72-
[https://github.com/nesaorg/Equivariant-Encryption-for-AI/blob/main/CONTEST.md](https://github.com/nesaorg/Equivariant-Encryption-for-AI/blob/main/CONTEST.md)
107+
[https://github.com/nesaorg/Equivariant-Encryption-for-AI/blob/main/CONTEST.md](https://github.com/nesaorg/Equivariant-Encryption-for-AI/blob/main/CONTEST.md)
File renamed without changes.

docs/nesa-logo.png

109 KB
Loading

docs/nesa-transparent-logo.png

36.8 KB
Loading
File renamed without changes.

0 commit comments

Comments
 (0)