Ghostor: Toward a Secure Data Sharing System from Decentralized Trust
summary
System hides which user accesses which file and lets clients detect stale or tampered reads using a verifiable anonymous history: per-object hash chains of signed digests where reader/writer signing keys are shared via anonymously distributed capabilities (keep-private encryption). Decentralized trust uses a blockchain one checkpoint per epoch (single hash) off the critical path. Concurrent writes use a two-phase prepare/commit to prevent time-stretch attacks. Prototype on EC2 reports ~4–5x throughput overhead versus insecure baseline (~25 ms extra server latency); Tor and PoW push client latency to seconds.
AI-written summary · about
chapters (18)
- 0:00Intro
- 0:18Motivating Example: Medical Record System
- 1:56Ghostor: Cryptographic Data Sharing System
- 2:13E2EE Data Sharing vs. Ghostor's Anonymity
- 3:18Verifiable Linearizability
- 4:05Comparison to Existing Work
- 4:44Bootstrapping Decentralized Trust
- 5:18Strawman: Use a Blockchain
- 5:56Ghostor's System Architecture
- 7:31Verifiable History (Strawman)
- 10:05Shared Capabilities
- 12:03Additional Challenge: Concurrent Operations
- 13:40This Technique Does Not Work for Writes
- 15:38Ghostor Stack
- 16:00Implementation
- 16:17Server-Side Latency to PUT a 1 MiB Object
- 17:10Total Latency
- 17:43Conclusion
related talks
19mCollective Authorities: Transparency & Decentralized Trust...
01h 01mKey transparency: Blockchain meets NoiseSocket
54mDistributed Trust: Is “Blockchain” the answer?
23mOblivP2P: An Oblivious Peer-to-Peer Content Sharing System
20mShared Security in the BLOCKCHAIN
42mWhy Privacy Matters
34mHyperjournal - A tamper-proof publishing system
12mThe Privacy Paradox: Balancing Transparency and Security in Blockchain