Angehefteter Tweet
Sunscreen
423 posts
Sunscreen
@SunscreenTech
Add Fully Homomorphic Encryption (FHE) to your app in minutes, not months.
encrypted Beigetreten Nisan 2021
16 Folgt21.6K Follower
Private information retrieval is the Trojan Horse of homomorphic encryption.
It's already deployed to hundreds of millions of users by Apple and Meta.
We broke down how it works and built a private DNS demo to show it in action: blog.sunscreen.tech/private-lookup…
English
Sunscreen retweetet
This is BIG news for FHE (Fully Homomorphic Encryption) adoption.
spectrum.ieee.org/fhe-intel
Tech being built by players like @SunscreenTech and @zama could one day be deployed at scale on specialized hardware like Heracles.
Now imagine if companies like @nvidia eventually enter the space with chips designed for encrypted computing. 🤯
We’re still so early.
English
Sunscreen retweetet
most people don't care about privacy, unless it enables a new technology or use case.
Naval@naval
Say something true, in an interesting way.
English
Sunscreen retweetet
The Achilles’ heel of software is its need to decrypt data for processing.
FHE eliminates that window of exposure, enabling what some call “full-privacy computing."
We are loving reading this from @barisozmen_twi
Read: bozmen.io/fhe
English
FHE vs TEEs
Read more: @trekeptor/tee-vs-fhe-two-paths-to-privacy-preserving-computation-67dd583732d7" target="_blank" rel="nofollow noopener">medium.com/@trekeptor/tee…
English
We're excited to announce our partnership with @latticaai to bring production-ready TFHE to developers.
Our Parasol compiler helps builders turn ordinary code into high‑performance FHE programs.
With Lattica's managed infrastructure, you can deploy encrypted computation without spinning up your own servers.
What this means: compile locally with Sunscreen, deploy to Lattica, and run programs on encrypted data in production. No plaintext exposure, no infrastructure headaches.
This opens the door for applications requiring exact computation on sensitive data - from smart contracts to secure voting to genomic analysis.
Get started with 3 free compute hours: lattica.ai/tfhe
Read more on our blog: blog.sunscreen.tech/deploying-tfhe…
English
Sunscreen retweetet
English
Sunscreen retweetet
Sunscreen retweetet
Private compute isn't just one thing, it's a spectrum of a trust model.
TEEs trust hardware, FHE trusts math.
Regardless of whether you're building an AI or blockchain app, you should ask: what (or who) are you trusting and whether that trust assumption works for your app.
English
Sunscreen retweetet
Spot on. AI needs private inference YESTERDAY.
Miguel de Vega@miguel_de_vega
Take ChatGPT for example. Everything you type: - can be used as evidence against you in court (there’s no legal confidentiality like you’d have with a therapist or lawyer): economictimes.indiatimes.com/tech/artificia… - can be used by default to train the model: openai.com/en-GB/policies… That’s what they’re telling you. What they’re not saying is the scary part: they’re basically building the most accurate model of you that’s ever existed: what you like, what you fear, what you’ll say yes to, what gets under your skin. And once someone has that, manipulating you gets way easier and way more precise than ads have ever been. The data you share is the power you give them to steer and manipulate you. Being safe is also staying free. Having your personal conversations with a chatbot like nilGPT (i.e. one that can cryptographically prove that nobody, not even Nillion, can access them) is a new kind of self-care. And honestly, a form of self-respect.
English
Sunscreen retweetet
Both TEEs and FHE enable privacy-preserving computation, but they approach the problem from different angles.
If performance is critical and trust in hardware is acceptable → TEE is a practical choice.
If absolute data privacy and decentralization matter most → FHE is the holy grail.
English
FHE lets you compute directly on encrypted data, without needing to decrypt data at all.
Benefits:
1/ Trust in math; cryptographic guarantees
2/ Good for cloud or decentralized systems where zero trust is important
3/ When data is never decrypted, the attack surface is reduced significantly.
Disadvantages:
1/ Currently slower than TEEs for many commercial use cases (but improving fast)
2/ Dev effort can be slightly higher.
English