Thread by @verumcapital

#AI and #ZK are 2023 hot narratives.

What if we tell you that those two will probably combine with ZK-Machine Learning projects?

A 🧵👇

ZK proofs allow for cryptographic proof of possession...without revealing the underlying data itself (check our ELI5 ZK thread for more info!).

The increased capabilities of zk-SNARKs have enabled exploration of new use cases.

AI Trustworthiness & accuracy will be one of them.

https://twitter.com/1051436010964439040/status/1625475175423647746

And lately, there has been a surge of interest in zk-SNARK enabled machine learning (ZK-ML) systems in the space.

You know it's serious stuff when Ethereum Foundation itself releases a video on that topic.

Have a look:

Why the need for ZK-ML? 👇

When using supervised machine learning, inputs are given to a model that has already been trained with specific parameters.

The model then produces an output that can be used by other systems.

Thanks to lightweight ML frameworks and formats like ONNX, it’s now possible to run these inferences on edge devices like phones or IoT devices, rather than sending the input data to centralized servers.

This improves both scalability and privacy for the users.

But what if:
→ We want to keep the input & parameters of ML models private?
→ The ML final user wants to verify that the input is trustworthy?

That's where Machine Learning X zkSNARK combination comes into play 👇

3 examples of how zk-SNARKs can be used for Machine Learning models:
→ Privacy-preserving machine learning
→ Verifiable machine learning
→ Secure machine learning

1/ PRIVACY-PRESERVING ML

🔹Zk-SNARKs can be used to train a ML model on private data without revealing that data to the model’s creators or users.

🔹Keyword: PRIVACY PRESERVATION

🔹Applications: sensitive/regulated industries (healthcare or finance...)

2/ VERIFIABLE ML

🔹Zk-SNARKs can be used to prove that a ML model was trained on a specific dataset, without revealing the details.

🔹Keyword: ML TRUST

🔹Applications: credit scoring, medical diagnosis

3/ SECURE ML

🔹Zk-SNARKs can be used to protect the integrity of ML models by ensuring that the model has not been tampered with/replaced with a different model.

🔹Keyword: ML SECURITY

🔹Applications: untrusted environments, such as edge devices or public clouds

What can be concrete USE CASES?

Leveraging ZK-ML oracles allows a simpler, faster & more efficient way to transfer offchain data to the blockchain.

ZK-ML could enable "smart judges" to interpret ambiguous events.

This will open the door to an unthinkable amount of use cases 👇

Non exhaustive ZK-ML use case list:

🔹ZK KYC: Person's identity X ID matching verification
🔹Fraud checks: ZK-ML fraud spam check for unusual behaviour
🔹Making DAOs Autonomous: enabling more efficient & accurate decision making, assessments & communication systems

CHALLENGES

ZK-ML integration is still very early stage. As a young technology, it faces many challenges:
🔹Regulatory
🔹ML private data (or model parameters) transparency
🔹ML models opacity/misbehavior
🔹Etc.

More information in this great resource:
0xparc.org/blog/zk-mnist

CONCLUSION

ZKP integration in AI systems will provide a new level of security, privacy and efficiency, allowing for powerful applications & use cases.

Definitely something to have an 👀 on, as we wouldn't be surprised to see future ZK X ML crypto projects coming soon!

Read our full article here:

medium.com/verum-capital-…

Tagging some CT gigabrains that could be interested:
@rektdiomedes
@Chinchillah_
@CryptoShiro_
@CryptoGirlNova
@crypthoem
@CJCJCJCJ_
@defi_mochi

@rektdiomedes @Chinchillah_ @CryptoShiro_ @CryptoGirlNova @crypthoem @CJCJCJCJ_ @defi_mochi That's all for this thread! If you 💙ed it:

1. Follow us at @verumcapital
2. Like and retweet the initial tweet:

https://twitter.com/1051436010964439040/status/1629080451632574466