hn.today

Sharded, encrypted storage between friends over Yggdrasil

github.com14 points4 comments
Screenshot of Sharded, encrypted storage between friends over Yggdrasil

A peer-to-peer, friend-group storage system built on the Yggdrasil overlay that shards and encrypts files across members' machines so no single failure loses data. Each file is encrypted with a fresh AES-256-GCM key, split into 4 MiB chunks, and each chunk erasure-coded into 6 shards (4 data + 2 parity). Shards are content-addressed by SHA-256 and distributed so a node never holds more than one shard per chunk (or two per host), with an option to mark slow nodes to limit their share. Uploaders produce a small stub (.ystub) that contains the manifest and the file key; sending a sealed stub to someone grants them access without a central cloud. Integrity is enforced by precomputed single-use challenges (20 per shard) that verify peers still hold data, and only the original writer can delete shards. Node identities derive from public keys (Yggdrasil addresses) and servers answer only listed peers (peers.json).

Features include versioned storage with delta saves, a gateway for S3-compatible access (paid customers), built-in Yggdrasil networking (no root or daemon required), messaging and topic delivery, and the ability to publish folders as highly available websites. Membership uses one-time invites (yggjoin1:…), admin-controlled peers.json pushed to nodes, and a dashboard for sharing via contact codes (ys1…). Releases and builds are available for major platforms (Go 1.24+), but the project is experimental and its cryptography has not been independently audited; do not rely on it as the sole copy of irreplaceable data.

Read on github.com4 comments on Hacker News

Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.

More in Security

Is Criminal Forensics Bullshit?

Is Criminal Forensics Bullshit?

Sara Gordon's research questions the scientific validity of forensic techniques like blood pattern analysis and ballistics, which are often accepted uncritically in court. Studies show these methods have significant uncertainties and rely more on subjective interpretation than rigorous science. (nautil.us)

The daily digest

Today's best Hacker News stories, summarized and screenshotted, one email a day.