SPVN

Scalable Payment Verification Network

Transcript

1. Scalable Payment Verification Network Tom Harding Bitcoin XT Maintainer Satoshi’s Vision Conference, 9/25/2016
2. Scale the Things We Know How to Scale Simplified Payment Verification Satoshi’s scaling solution Millions of SPV wallets in use every day Scalable Payment Verification Network Improved SPV security Build confidence in unconfirmed transactions, which SPV users rely on New capabilities enable Light Network Nodes Let anyone contribute meaningfully to security Allow the blockchain to grow faster No consensus changes Completely supplemental
3. SPV: Satoshi’s Decentralized Scaling Solution Wallet stores a highly pruned blockchain (headers, personal transactions, and merkle proofs) • Security derives from confirmations • Talks to bitcoin nodes, not proprietary servers • User has full control of his keys
4. SPV: Vulnerabilities • The mining/fullnode network could hard fork without you knowing • Your network provider could omit information SPV called “toxic light client” and compared to mercury poisoning, and playing with fire • Privacy concerns related to bloom filters The market has accepted these risks. Millions of SPV wallets are in use every day. But we can make it more secure and more scalable
5. SPV: Satoshi’s Fraud Proofs • Briefly described as “one strategy” in bitcoin.pdf ◦ Receive alerts when a remote node detects an invalid block ◦ Suggestion is to “download the full block” but… ▪ An SPV client can’t validate a full block ▪ We’d like to support larger blocks ◦ Compact block invalidity (“fraud”) proofs would let a wallet detect a hard fork Such detection is not the subject of this talk ◦ How do we build a Scalable Payment Verification Network?
6. Improving SPV Security Immediately anchor received payment confidence to the blockchain • Fetch input proofs ◦ Absence of a positive input proof can be used to distrust a payment • Fetch spentness proofs and receive spend alerts ◦ Absence of a spentness proof can be used build confidence probabilistically • Add more complete local transaction validation rules Improved network practices • Split filters among peers • Peer rotation • Enable (but don’t require) use of trusted peers • Explore use of tor to reach the P2P network
7. Improving SPV Security Wallets need everything on the diagram in the next slide. Today, they only receive... The received payment transaction Transactions spending M
9. Original Bitcoin Network
10. Today’s Bitcoin Network
11. New Capabilities for Light Network Nodes SPV-serving nodes need only SPV security themselves Capabilities (a la carte): • Prover Provide input/spentness proofs for stem • Prover Finder Finds provers for stem • Transaction Repeater Host bloom filters and relay transactions
12. • Provides input/spentness proofs for a TXID stem, like 0x828fea • Stores and maintains a blockchain shard set by TXID stem ◦ Size of slice is completely configurable ◦ Also tracks unconfirmed transactions in the set • Registers stems served with prover finders • Collectively, the network can validate anything, falling back to full nodes • Light nodes decentralize the collective validation function Prover
13. • Finds provers for TXID stem • Resolution service very similar to DNS required • Hierarchical TXID name space partition with full nodes at root • Actually using DNS is a likely solution Prover Finder
14. • Hosts bloom filters • Relays transactions ◦ Achieve “SPV” security by checking for relay from multiple full nodes • Scales bloom filter hosting • Decentralizes bloom filter hosting • Is a prover finder client Transaction Repeater
15. Scalable Payment Verification Network
16. Properties of SPVN Increased TX rate with maintained decentralization Increased light wallet security No consensus changes required No changes to transaction authoring or acceptance No capital funding required per wallet hosted No co-signers necessary Lots of engineering work to be done Can be built incrementally Prover can be emulated on mainnet using full nodes Not actually needed until TX rate has increased well above 1MB/10min