東京へようこそ Welcome to Tokyo!

Transcription

1 東京へようこそ Welcome to Tokyo!

2 Siddharth Singhal Project Ubin

3 Agenda 1. Introduction to CBDCs 2. Cash on Corda 3. Project Ubin: Introduction and key observations 4. Demo and Decentralised LSMs

4 1. CBDC Introduction

5 A wholesale payments solution on blockchain 2018 is the year of production for blockchains Systemic value of blockchains can only be realised with cash on-ledger Cryptocurrencies and private sector e-cash initiatives have first mover advantage, but reduce transparency for Central Banks. R3 has a suitable solution for Central Banks to facilitate wholesale payments between known participants and advantage of blockchain R3 has a proven track record with groups of Banks and Central Banks

6 Framing the topic: Wholesale digital fiat on blockchains! We are not suggesting issuing central bank currency in digital format to households -Unanswered questions around intrinsic interest provision, anonymity/identity, scalability, risks in times of systemic uncertainty, etc -See Appendix We are discussing the merits of recording the ownership of reserves on distributed ledgers (blockchains) - Participants are existing RTGS participants ie clearing banks etc - Participation could be extended stepwise to other financial intermediaries and a whitelist of corporates

7 The business case per jurisdiction Central Bank mandate: Mandates differ by jurisdiction, but a common goal is maintaining an orderly financial ecosystem, systemic stability, and responding to technological advances Tokenised wholesale currency will: Reduce systemic central-point-of-failure risks Reduce Herstatt-related risks Add value to the state s ecosystem, fully enabling the promise of blockchains Drive increased velocity of business within the ecosystem Prepare the state for the next step in the evolution of payment systems

8 2. Corda Cash on Corda

9 R3: Progress made with Central Banks Jasper (Bank of Canada) - Phase 1: Ethereum, Monetary model, PFMIs - Phase 2: Corda, LSM - Phase 3: DvP Ubin (Monetary Authority of Singapore) - Phase 1: Ethereum, Monetary model (Pledge, transfer, redeem) - Phase 2: Corda, Fabric, Quorum (RTGS, LSM) - Next phases: Bonds and DvP Cross border payments including FX Legal basis LionRock (Hong Kong Monetary Authority) - Phase 1: Corda, RTGS, Corporate access, DvP, atomic coupons - Next phases: DvP Cross border

10 Unrivaled work on Cash across the DLT landscape Blog posts Papers Decks Research IP Pledge/Issue/Tx/Redeem model Liquidity Saving Mechanisms Jasper Ubin LionRock Projects Argent Pilots Partners Systems Integrators Law firms Opportunities Central banks Commercial banks

11 3.1 Ubin Introduction

12 Project Ubin Phase2 Overview Collaborative design and rapid prototyping project, exploring the use of Distributed Ledger Technologies (DLT) for Real-Time Gross Settlement (RTGS) System Led by the Monetary Authority of Singapore (MAS) and The Association of Banks in Singapore (ABS). It is managed and delivered by Accenture, with a consortium of 11 financial institutions in Singapore and 4 technologies partners 12 The 13 weeks project successfully demonstrated that RTGS functions, especially Liquidity Saving Mechanism may be decentralised without comprising privacy 3 prototypes were developed using three DLT platforms: Corda, Hyperledger Fabric and Quorum Ref:

13 UBIN 2 KEY REQUIREMENTS Cash issuance/payment/destruction Queuing Mechanism Data privacy Deterministic settlement finality Liquidity-saving mechanism (LSM)

14 CASH Cash is an own able fact. Ownership can be transferred. Cash Owner: Alice Currency: SGD Amount: $10 8 Cash Owner: Bob Currency: SGD Amount: $10 Alice paid $10

15 QUEUEING MECHANISM An Obligation is a shared fact. Obligations will be settled by transferring cash at a future point in time. ALICE Id From To Amt By Penalty Paid 1 Alice Bob $ % daily $0 BOB Id From To Amt By Penalty Received 1 Alice Bob $ % daily $0 Alice Owes Bob $10 Payable by 31/03/17 9

16 PRIVACY A B No broadcast, need to know, uses point to point communication Confidential Identities, shared on a need to know basis D C

17 DETERMINISTIC SETTLEMENT FINALITY A transaction is an indivisible and irreducible set of changes such that either all occur, or nothing occurs. O 0 O 1 O 1 O 2 C 0 C 0 C 0 C 0 11

18 CONSENSUS - Service Corda decouples the consensus mechanism from the ledger. This unique capability is what enables an open network of nodes that can freely connect while retaining privacy. Notary is our term for this group of servers operated in a consensus pool providing a uniqueness service.

19 CONSENSUS - Pluggable Consensus is pluggable at transaction level, not system level Consensus algorithm tailored to needs of transaction type Currently implemented RAFT and BFT

20 APPROACH TO UBIN 2 REQUIREMENTS Key Requirements Cash issuance/payment/destruction Cash contract Approach Queuing Mechanism Obligation contract Data privacy Data propagated only to relevant nodes Deterministic settlement finality Transactions are atomic Liquidity-saving mechanism (LSM) Range of innovative options

21 3.2 Ubin Key Observations

22 MODELLING THE PROBLEM All payments are modelled as obligations (like IOUs) Payments are all made against an obligation Each node prioritises it s own obligations Each node chooses how much liquidity it will offer towards netting operations

23 CHALLENGES We don t want to reveal everyone s unsettled payments We need fast convergence We can t stop the world We must never be left in an incomplete state!

24 KEY OBSERVATIONS & FINDINGS PRIVACY 1 Privacy is inherent Need to know, point-to-point, instead of global broadcast 2 Shielding participants identities Confidential Identities: unique pair of public key to be exchanged between the sender and receiver for every transaction 3 In the Ubin Phase 2 prototype, transaction amount was not shielded. Among the pluggable algorithm choices, a cycle-solver was used (Details in LSM section) In any production implementation, a relevant design choice can be implemented for LSM algorithm

25 KEY OBSERVATIONS & FINDINGS SCALABILITY & PERFORMANCE 1 Multiple notary services or cluster required To allow load balancing and avoid single point of failure 2 Transaction graph can be trimmed through re-issuance of the assets Provenance chain of an asset can be long after several handovers of an asset. 3 Cycle-solver does not require the network to freeze Payment instructions can be processed while gridlock resolution is running

26 KEY OBSERVATIONS & FINDINGS RESILIENCY 1 Network continue operates without dependency to any participant 2 In this prototype, single notary service can be improved Cluster consensus services can be implemented and operated by multiple parties to support geographically diverse transacting parties

27 KEY OBSERVATIONS & FINDINGS FINALITY 1 Notary ensures finality By obtaining a notary signature, participants of the transaction can be equally sure that the input states are unconsumed (unspent) by prior transactions 2 Atomic transaction in gridlock settlement Settlement transaction is executed through a single atomic netting transaction

28 CHALLENGES Fulfilled We don t want to reveal everyone s unsettled payments We need fast convergence We can t stop the world We must never be left in an incomplete state!

29 Road ahead on Ubin

30 4.1 Demo Core Functions

31 GRIDLOCK RESOLUTION NODE MAP Five banks will be used to walkthrough Gridlock Resolution: C M Credit Suisse MUFG Credit Suisse MAS Regulator MAS Central Bank MUFG MEPS+ Simulated with an interface stub (for Pledge and Redeem) J J.P. Morgan O OCBC S Standard Chartered J.P. Morgan OCBC Standard Chartered 6 other banks

32 CORE FUNCTIONS DEMO J.P. MORGAN CREDIT SUISSE Pledge $1,000,000 $0 $0 MAS MEPS+ PLEDGE Approve Pledge $1,000,000 $1,000,000 $0

33 Demo Video - Pledge

34 CORE FUNCTIONS DEMO J.P. MORGAN CREDIT SUISSE Pledge $1,000,000 $0 $0 MAS MEPS+ PLEDGE Approve Pledge $1,000,000 $1,000,000 $0 Transfer $700,000 (Normal Priority) FUND TRANSFER $300,000 $700,000

35 Demo Video - Transfer

36 CORE FUNCTIONS DEMO J.P. MORGAN CREDIT SUISSE Pledge $1,000,000 $0 $0 MAS MEPS+ PLEDGE Approve Pledge $1,000,000 $1,000,000 $0 Redeem $190,000 $300,000 Transfer $700,000 (Normal Priority) $700,000 FUND TRANSFER MAS MEPS+ Approve Redeem $190,000 $110,000 $700,000 REDEEM

37 Demo Video - Redeem

38 CORE FUNCTIONS DEMO J.P. MORGAN CREDIT SUISSE $110,000 $700,000 Transfer $150,000 (Normal Priority) Transfer $320,000 (Normal Priority) Transfer $280,000 (Normal Priority) QUEUE QUEUE From To Amount Priority Status J.P. MORGAN CREDIT SUISSE $150,000 Normal Active J.P. MORGAN CREDIT SUISSE $320,000 Normal Active J.P. MORGAN CREDIT SUISSE $280,000 Normal Active

39 Demo Video - Queue

40 CORE FUNCTIONS DEMO J.P. MORGAN $110,000 CREDIT SUISSE $700,000 QUEUE Transfer $150,000 (Normal Priority) Transfer $320,000 (Normal Priority) Transfer $280,000 (Normal Priority) From To Amount Priority Status J.P. MORGAN CREDIT SUISSE $150,000 Normal Active J.P. MORGAN CREDIT SUISSE $320,000 Normal Active J.P. MORGAN CREDIT SUISSE $280,000 Normal Active Update Priority $280,000 HIGH PRIORITY QUEUE QUEUE From To Amount Priority Status J.P. MORGAN CREDIT SUISSE $280,000 High Active J.P. MORGAN CREDIT SUISSE $150,000 Normal Active J.P. MORGAN CREDIT SUISSE $320,000 Normal Active UPDATE PRIORITY $110,000 $700,000

41 Demo Video - High Priority

42 CORE FUNCTIONS DEMO J.P. MORGAN $110,000 CREDIT SUISSE $700,000 Update Status $150,000 ON-HOLD QUEUE From To Amount Priority Status J.P. MORGAN CREDIT SUISSE $280,000 High Active J.P. MORGAN CREDIT SUISSE $150,000 Normal On-Hold J.P. MORGAN CREDIT SUISSE $320,000 Normal Active UPDATE STATUS ON-HOLD $110,000 $700,000

43 Demo Video On Hold

44 CORE FUNCTIONS DEMO J.P. MORGAN CREDIT SUISSE Pledge $350,000 Approve Pledge $350,000 MAS MEPS+ $110,000 QUEUE $110,000 $460,000 QUEUE Update Status: $150,000 ON-HOLD From To Amount Priority Status J.P. MORGAN CREDIT SUISSE $280,000 High Active J.P. MORGAN CREDIT SUISSE $150,000 Normal On-Hold J.P. MORGAN CREDIT SUISSE $320,000 Normal Active Settle Queues From To Amount Priority Status J.P. MORGAN CREDIT SUISSE $150,000 Normal On-Hold J.P. MORGAN CREDIT SUISSE $320,000 Normal Active $700,000 $700,000 $700,000 UPDATE STATUS ON-HOLD SETTLE QUEUE UPON SUFFICIENT LIQUIDITY $180,000 $980,000

45 Demo Video Pledge, Settle Queue

46 CORE FUNCTIONS DEMO J.P. MORGAN $180,000 CREDIT SUISSE $980,000 Cancel $150,000 QUEUE From To Amount Priority Status J.P. MORGAN CREDIT SUISSE $150,000 Normal On-Hold J.P. MORGAN CREDIT SUISSE $320,000 Normal Active CANCEL $180,000 $980,000

47 Demo Video Cancel

48 CORE FUNCTIONS DEMO J.P. MORGAN $180,000 CREDIT SUISSE $980,000 Cancel $150,000 QUEUE From To Amount Priority Status J.P. MORGAN CREDIT SUISSE $150,000 Normal On-Hold J.P. MORGAN CREDIT SUISSE $320,000 Normal Active CANCEL $180,000 Update Status: $150,000 ACTIVATE $980,000 QUEUE From To Amount Priority Status J.P. MORGAN CREDIT SUISSE $150,000 Normal Active UPDATE STATUS RE-ACTIVATE No Transaction in Queue Settle Queues $30,000 Settle Queues $1,130,000

49 Demo Video Cancel, Reactivate, Settle Queue

50 4.2 Decentralised LSMs A walkthrough of the approach in UBIN- Ph2 prototype

51 GRIDLOCK RESOLUTION NODE MAP Five banks will be used to walkthrough Gridlock Resolution: C M Credit Suisse MUFG Credit Suisse MAS Regulator MAS Central Bank MUFG MEPS+ Simulated with an interface stub (for Pledge and Redeem) J J.P. Morgan O OCBC S Standard Chartered J.P. Morgan OCBC Standard Chartered 6 other banks

52 GRIDLOCK RESOLUTION SCENARIOS GRIDLOCK SCENARIO 1 GRIDLOCK SCENARIO 2 T1=5k C (3k) T5=7k T1=5k C (3k) T8=8K M (4k) S (3k) M (4k) T9=100k T7=6k T10=5k S (3k) T2=6k T4=8k T2= 6k T6=7k J (5k) T3=7k O (4k) T3=30k J (5k) T4= 8k O (4k) T5=80k

53 TWO STAGE APPROACH

54 DETECT Any node may initiate a scan, but let us first consider it being started by node A

55 DETECT A initiates the scan and requests nodes B and C continue it. A records edge AB and AC.

56 DETECT Node B subsequently requests that node C continue the scan, while node C requests both nodes B and D continue.

57 DETECT Node D responds with the details of the edge DC. At this point it is possible that either node B or C may take different actions, dependent on the timing of the scan, but node C bundles the reply from D regarding edge DC, while adding details of the edge CA. Node B offers details of the edge BC.

58 DETECT Node A collates the replies from B and C, forming a view of a graph with edges AB, CA, BC and DC.

59 PLAN Find cycles from detect phase Prioritise cycles Compute netting payments Agree netting solution with all participants Execute atomic netting payments and settlements Netting solutions are always cycles!

60 PLAN (1 minute overview) 1. Split graph into strongly connected component (Tarjan) 2. Find all cycles in graph (Johnson) 3. Overlay obligations onto each cycle 4. Consider whether each cycle can be netted 5. Pick the cycle that meets some pre-defined heuristic Largest gross amount of obligations settled Largest cycle 6. Return netting payments and obligations to settle

61 EXECUTE Consider the case where C needs to make a payment to B to settle ABCA. A plans the netting solution and finds C must pay B. Generates a transaction to make one payment that settles the three payment obligations.

62 EXECUTE A builds a payload with the anonymised netting details, adds his obligations AB and AC.

63 EXECUTE As C is a payee, they generate a new anonymous public key that B should make a payment to. C can also add the edge BC to the transaction builder. C then sends the payload to B.

64 EXECUTE B makes a payment to the anonymous key added by C. There are no more obligations to add as A and C have already added them. The collection of states is complete, so B can now sign this transaction proposal B then sends the proposal to A for signing.

65 EXECUTE NOTARY CLUSTER A signs then sends to C, who also signs. C then sends to B. At this point all the signatures have been collected. B then sends the fully signed transaction to the notary for finalisation. B can this distribute the finalised atomic transaction to A and C by doing one more circumnavigation of the cycle.

66 Thank you