LibreSocial · Documentation

Access Control & Security

Deep-dive page: how a public, untrusted DHT stores private social data — identity, encryption, ACLs, group keys, revocation — and what remains open.

Supersession note: the identity/ACL baseline is LCN 2009; DDS-level security (buckets, groups, consolidation) is CTS 2014; the authoritative access-control model is the short paper LANMAN 2017 (poster variant: same venue). Crypto parameters in the 2009 paper (RSA-1024) are superseded by the elliptic-curve implementation in com.p2pframework.security.ellipticcurves; see security component for code state.

Identity: the public key is the identifier

Registration (LCN 2009):

Hybrid encryption of every stored object

Groups and hierarchy

(CTS 2014):

Write protection for distributed data structures

Two-stage write control (CTS 2014):

  1. Bucket signature under the group key → who may insert into a bucket;
  2. element signature + stored nonce → only the author may modify/delete that element, with replay protection.

Lists live in buckets (bucketKey = h(listName + floor(i/splitsize))), remote operations (get(i), set(i), contains) run at the storage peer; deletion inserts a synthetic deletedItem and a consolidation protocol shifts it exactly one bucket per invocation (two signatures, two storage peers, notification timeout) so nothing gets lost. Simulation: deleted-item share falls from up to 0.3 to mostly < 0.05; splitsize ≈ 20 was a good compromise. Details: DDS internals.

Measured overhead (2009, Core 2 Quad — still the documented cost model)

Operation Cost
Secure message < 2 KB overhead (1899–1911 B absolute); relative 212 % (895 B payload) → 17 % (10.9 kB)
En-/decrypt message ~10–14 ms / 8–10 ms
Each extra authorized user +413 B (key-list copy)
200 privileged users 83,653 B (24 % of a 347 kB object), encrypt 89 ms, decrypt ~20 ms constant
Key wrapping 0.36 ms per key

Documented guidance: from ~200 friends upward, use group-based access control instead of per-friend key copies.

Known weak points and 2026 open work

Flagged by the researchers themselves (work items → security roadmap blog):

Sources