SLIDE 1
Sergi Delgado-Segura, Surya Bakshi, Cristina Pérez-Solà, James Litton, Andrew Pachulski, Andrew Miller and Bobby Bhattacharjee
TxProbe: Discovering Bitcoin’s Network Topology Using Orphan Transactions
sr_gi
#SFBW19
WHAT DO WE KNOW ABOUT THE TOPOLOGY? WHAT DO WE KNOW ABOUT THE - - PowerPoint PPT Presentation
WHAT DO WE KNOW ABOUT THE TOPOLOGY? WHAT DO WE KNOW ABOUT THE - - PowerPoint PPT Presentation
# SFBW19 TxProbe: Discovering Bitcoins Network Topology Using Orphan Transactions Sergi Delgado-Segura , Surya Bakshi, Cristina Prez-Sol, James Litton, Andrew Pachulski, Andrew Miller and Bobby Bhattacharjee sr_gi WHAT DO WE KNOW ABOUT THE
SLIDE 2
SLIDE 3
WHAT DO WE KNOW ABOUT THE TOPOLOGY?
Number of nodes and location of them
SLIDE 4
WHAT DO WE KNOW ABOUT THE TOPOLOGY?
Number of nodes and location of them
SLIDE 5
WHAT DO WE KNOW ABOUT THE TOPOLOGY?
Number of nodes and location of them The edges are hidden by design
SLIDE 6
WHY HAVE A HIDDEN TOPOLOGY?
An open topology could ease different types of attacks:
- Transaction deanonymization
- Network based attacks (e.g: Eclipse attacks)
The current approach of the Bitcoin Core is to keep it hidden
SLIDE 7
WHY HAVE AN OPEN TOPOLOGY?
We know nothings about how the network really is:
- Is the network decentralised?
- Are there supernodes controlling the network traffic?
- Information withholding
- Censorship
- Are there weak spots in the network that can be easily isolated?
Security by obscurity does not seem the proper way to go
SLIDE 8
THE TOPOLOGY SHOULD LOOK RANDOM
How Bitcoin (Core client) nodes choose their peers?
- Pseudorandomly from the addrman
- 8 outbound connections by default
No pair of nodes in the same /16 (IPv4)
- 117 inbound connection by default (no IP restriction here)
Bitcoin forks based on the Core client follow the same approach
SLIDE 9
BACKGROUND
Our inferring technique is based on transaction propagation We take advantage of how transactions are handled by nodes:
- orphans transactions
- double-spending transactions
SLIDE 10
Valid transaction are stored in mempool Transaction in mempool are eventually propagated throughout the node neighborhood
A B
TRANSACTION PROPAGATION IN BITCOIN
SLIDE 11
Valid transaction are stored in mempool Transaction in mempool are eventually propagated throughout the node neighborhood
A B i n v ( h ( t x
n
) ) a n n
- u
n c e
TRANSACTION PROPAGATION IN BITCOIN
SLIDE 12
Valid transaction are stored in mempool Transaction in mempool are eventually propagated throughout the node neighborhood
A B i n v ( h ( t x
n
) ) get_data(h(txn)) a n n
- u
n c e request
TRANSACTION PROPAGATION IN BITCOIN
SLIDE 13
Valid transaction are stored in mempool Transaction in mempool are eventually propagated throughout the node neighborhood
A B i n v ( h ( t x
n
) ) get_data(h(txn)) a n n
- u
n c e request
TRANSACTION PROPAGATION IN BITCOIN
Wait for tx up to 2 min
SLIDE 14
Valid transaction are stored in mempool Transaction in mempool are eventually propagated throughout the node neighborhood
A B inv(h(txn)) get_data(h(txn)) tx(txn) announce request deliver
TRANSACTION PROPAGATION IN BITCOIN
Wait for tx up to 2 min
SLIDE 15
A transaction is orphan if some of the referenced UTXOs are unknown They can not be validated, so they are stored in a separated data structure known as MapOrphanTransactions (or OrphanPool for short) Transactions in MapOrphanTransactions are NOT forwarded to any node If the same transactions is offered again to the node (inv message), it will not requested back (getaddr)
ORPHAN TRANSACTIONS
txB txC txD Orphan transaction
SLIDE 16
DOUBLE-SPENDING TRANSACTIONS
A
id = 4F3…ED
Source: 4F3…ED
To: Bob
Source: 4F3…ED
To: Alice
txB txB’
B
. . .
B’s mempool
txn-1 tx0
SLIDE 17
DOUBLE-SPENDING TRANSACTIONS
A
id = 4F3…ED
Source: 4F3…ED
To: Bob
Source: 4F3…ED
To: Alice
txB txB’
B
. . .
B’s mempool
txn-1 tx0
SLIDE 18
DOUBLE-SPENDING TRANSACTIONS
A
id = 4F3…ED
Source: 4F3…ED
To: Bob
Source: 4F3…ED
To: Alice
txB txB’
B
. . .
B’s mempool
txn-1 tx0 txB
SLIDE 19
DOUBLE-SPENDING TRANSACTIONS
A
id = 4F3…ED
Source: 4F3…ED
To: Bob
Source: 4F3…ED
To: Alice
txB txB’
B
. . .
B’s mempool
txn-1 tx0 txB
SLIDE 20
DOUBLE-SPENDING TRANSACTIONS
A
id = 4F3…ED
Source: 4F3…ED
To: Bob
Source: 4F3…ED
To: Alice
txB txB’
B
. . .
B’s mempool
txn-1 tx0 txB
SLIDE 21
A BASIC TOPOLOGY INFERRING TECHNIQUE
Two nodes Three transactions Observation tool (like coinscope)
SLIDE 22
A BASIC TOPOLOGY INFERRING TECHNIQUE
A
B
Two nodes Three transactions Observation tool (like coinscope)
SLIDE 23
A BASIC TOPOLOGY INFERRING TECHNIQUE
A
B
Two nodes Three transactions Observation tool (like coinscope)
txP txM Parent tx Marker tx txF Flood tx
id = 4F3…ED
SLIDE 24
A BASIC TOPOLOGY INFERRING TECHNIQUE
A
B
Two nodes Three transactions Observation tool (like coinscope)
txP txM Parent tx Marker tx txF Flood tx
id = 4F3…ED
US
SLIDE 25
POSITIVE INFERRING TECHNIQUE
A B
SLIDE 26
POSITIVE INFERRING TECHNIQUE
A B
US
SLIDE 27
POSITIVE INFERRING TECHNIQUE
A B
US
SLIDE 28
∅ ∅
A’s Mempool B’s Mempool
POSITIVE INFERRING TECHNIQUE
A B
∅
B’s MapOrphanTransactions
US
SLIDE 29
∅ ∅
A’s Mempool B’s Mempool
POSITIVE INFERRING TECHNIQUE
A B
∅
B’s MapOrphanTransactions txP txF (1) ( 1 )
US
SLIDE 30
∅ ∅
A’s Mempool B’s Mempool
POSITIVE INFERRING TECHNIQUE
A B
∅
B’s MapOrphanTransactions
US
SLIDE 31
∅ ∅
A’s Mempool B’s Mempool
POSITIVE INFERRING TECHNIQUE
A B
∅
B’s MapOrphanTransactions txP txF (1) (1)
US
SLIDE 32
txP txF
POSITIVE INFERRING TECHNIQUE
A B
∅ ∅
A’s Mempool B’s Mempool
∅
B’s MapOrphanTransactions txP txF (1) (1)
US
SLIDE 33
txP txF
POSITIVE INFERRING TECHNIQUE
A B
∅ ∅
A’s Mempool B’s Mempool
∅
B’s MapOrphanTransactions txP txF (1) (1)
US
SLIDE 34
POSITIVE INFERRING TECHNIQUE
A B
∅ ∅
A’s Mempool B’s Mempool
∅
B’s MapOrphanTransactions txP txF (1) (1)
US
SLIDE 35
POSITIVE INFERRING TECHNIQUE
A B
∅ ∅
A’s Mempool B’s Mempool
∅
B’s MapOrphanTransactions txP txF (1) (1) txM (2)
US
SLIDE 36
POSITIVE INFERRING TECHNIQUE
A B
∅ ∅
A’s Mempool B’s Mempool
∅
B’s MapOrphanTransactions txP txF (1) (1)
US
SLIDE 37
POSITIVE INFERRING TECHNIQUE
A B
∅ ∅
A’s Mempool B’s Mempool
∅
B’s MapOrphanTransactions txP txF (1) (1) txM (2) (2)
US
SLIDE 38
POSITIVE INFERRING TECHNIQUE
A B
∅ ∅
A’s Mempool B’s Mempool
∅
B’s MapOrphanTransactions txP txF (1) (1) txM (2) (2) txM (3)
US
SLIDE 39
POSITIVE INFERRING TECHNIQUE
A B
∅ ∅
A’s Mempool B’s Mempool
∅
B’s MapOrphanTransactions txP txF (1) (1) txM (2) (2)
US
SLIDE 40
POSITIVE INFERRING TECHNIQUE
A B
∅ ∅
A’s Mempool B’s Mempool
∅
B’s MapOrphanTransactions txP txF (1) (1) txM (2) (2) txM (3)
US
SLIDE 41
∅ ∅
A’s Mempool B’s Mempool
NEGATIVE INFERRING TECHNIQUE
A B
∅
B’s MapOrphanTransactions
US
SLIDE 42
∅ ∅
A’s Mempool B’s Mempool
NEGATIVE INFERRING TECHNIQUE
A B
∅
B’s MapOrphanTransactions
US
txP txF (1) ( 1 )
SLIDE 43
∅ ∅
A’s Mempool B’s Mempool
NEGATIVE INFERRING TECHNIQUE
A B
∅
B’s MapOrphanTransactions
US
SLIDE 44
∅ ∅
A’s Mempool B’s Mempool
NEGATIVE INFERRING TECHNIQUE
A B
∅
B’s MapOrphanTransactions txP txF (1) (1)
US
SLIDE 45
A B
NEGATIVE INFERRING TECHNIQUE
∅ ∅
A’s Mempool B’s Mempool
∅
B’s MapOrphanTransactions txP txF (1) (1) txM (2)
US
SLIDE 46
A B
NEGATIVE INFERRING TECHNIQUE
∅ ∅
A’s Mempool B’s Mempool
∅
B’s MapOrphanTransactions txP txF (1) (1)
US
SLIDE 47
A B
NEGATIVE INFERRING TECHNIQUE
∅ ∅
A’s Mempool B’s Mempool
∅
B’s MapOrphanTransactions txP txF (1) (1) txM (2)
US
SLIDE 48
US US
txM (2)
Positive edge inferring Negative edge inferring
(1) (3) (1)
A BASIC TOPOLOGY INFERRING TECHNIQUE
(1) (2) (1)
SLIDE 49
US US
txM (2)
Positive edge inferring Negative edge inferring
(1) (3) (1)
A B inv(h(txM))
US
A BASIC TOPOLOGY INFERRING TECHNIQUE
(1) (2) (1)
SLIDE 50
US US
txM (2)
Positive edge inferring Negative edge inferring
(1) (3) (1)
A B inv(h(txM))
US
A BASIC TOPOLOGY INFERRING TECHNIQUE
(1) (2) (1)
SLIDE 51
US US
txM (2)
Positive edge inferring Negative edge inferring
(1) (3) (1)
A B inv(h(txM))
US
A BASIC TOPOLOGY INFERRING TECHNIQUE
get_data(h(txM))
(1) (2) (1)
SLIDE 52
US US
txM (2)
Positive edge inferring Negative edge inferring
(1) (3) (1)
A B inv(h(txM))
US
A BASIC TOPOLOGY INFERRING TECHNIQUE
edge does not exist
get_data(h(txM))
(1) (2) (1)
SLIDE 53
US US
txM (2)
Positive edge inferring Negative edge inferring
(1) (3) (1)
A B inv(h(txM))
US
A BASIC TOPOLOGY INFERRING TECHNIQUE
(1) (2) (1)
SLIDE 54
US US
txM (2)
Positive edge inferring Negative edge inferring
(1) (3) (1)
A B inv(h(txM))
US
A BASIC TOPOLOGY INFERRING TECHNIQUE
∅
(1) (2) (1)
SLIDE 55
US US
txM (2)
Positive edge inferring Negative edge inferring
(1) (3) (1)
A B inv(h(txM))
US
A BASIC TOPOLOGY INFERRING TECHNIQUE
∅
edge does exist
(1) (2) (1)
SLIDE 56
ITS NOT THAT EASY
A B
Coinscope
US
SLIDE 57
ITS NOT THAT EASY
A B
Coinscope
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 58
ITS NOT THAT EASY
A B
Coinscope
C
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 59
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
ITS NOT THAT EASY
A B
Coinscope
C
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 60
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
ITS NOT THAT EASY
A B
Coinscope
C
t x P (1) txF (1)
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 61
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
ITS NOT THAT EASY
A B
Coinscope
C
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 62
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
txP
(1)
txF
(1)
ITS NOT THAT EASY
A B
Coinscope
C
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 63
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
txP
(1)
txF
(1)
ITS NOT THAT EASY
A B
Coinscope
C
t x P (2)
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 64
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
txP
(1)
txF
(1)
ITS NOT THAT EASY
A B
Coinscope
C
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 65
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
txP
(1)
txF
(1)
txP
(2)
ITS NOT THAT EASY
A B
Coinscope
C
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 66
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
txP
(1)
txF
(1)
txP
(2)
ITS NOT THAT EASY
A B
Coinscope
C
t x M (3)
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 67
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
txP
(1)
txF
(1)
txP
(2)
ITS NOT THAT EASY
A B
Coinscope
C
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 68
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
txP
(1)
txF
(1)
txP
(2)
ITS NOT THAT EASY
A B
Coinscope
C
txM
(3)
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 69
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
txP
(1)
txF
(1)
txP
(2)
ITS NOT THAT EASY
A B
Coinscope
C
txM
(3)
t x M (4)
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 70
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
txP
(1)
txF
(1)
txP
(2)
ITS NOT THAT EASY
A B
Coinscope
C
txM
(3)
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 71
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
txP
(1)
txF
(1)
txP
(2)
ITS NOT THAT EASY
A B
Coinscope
C
txM
(3)
txM
(4)
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 72
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
txP
(1)
txF
(1)
txP
(2)
ITS NOT THAT EASY
A B
Coinscope
C
txM
(3)
txM
(4)
txM (5)
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 73
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
txP
(1)
txF
(1)
txP
(2)
ITS NOT THAT EASY
A B
Coinscope
C
txM
(3)
txM
(4)
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 74
∅ ∅ ∅ ∅
A’s Mempool C’s Mempool B’s Mempool B’s MapOrphanTransactions
txM
(5)
txP
(1)
txF
(1)
txP
(2)
ITS NOT THAT EASY
A B
Coinscope
C
txM
(3)
txM
(4)
Long story short, the technique will fail if we add an additional node to the picture
US
SLIDE 75
Isolation
MAKE THIS WORK IN A REAL NETWORK
Synchrony Efficiency
A
txP
A
txP
B
txF
≈ O(n)
≈ O( n)
n = #nodes
SLIDE 76
Isolation
MAKE THIS WORK IN A REAL NETWORK
Synchrony Efficiency
A
txP
A
txP
B
txF
≈ O(n)
≈ O( n)
n = #nodes
SLIDE 77
t x P (2)
ISOLATION
A B
Coinscope
C
US
SLIDE 78
t x P (2)
ISOLATION
A B
Coinscope
C
US
SLIDE 79
t x P (2)
ISOLATION
A B
Coinscope
C
US
A B
C
SLIDE 80
t x P (2)
ISOLATION
A B
Coinscope
C
US
A B
C
INV
SLIDE 81
t x P (2)
ISOLATION
A B
Coinscope
C
US
A B
C
GETDATA INV
SLIDE 82
t x P (2)
ISOLATION
A B
Coinscope
C
US
A B
C
GETDATA TX INV
SLIDE 83
t x P (2)
ISOLATION
A B
Coinscope
C
US
A B
C
GETDATA TX INV
SLIDE 84
t x P (2)
ISOLATION
A B
Coinscope
C
US
A B
C
GETDATA TX INV
SLIDE 85
t x P (2)
ISOLATION
A B
Coinscope
C
US
A B
C
GETDATA TX INV
HOW?
SLIDE 86
t x P (2)
ISOLATION
A B
Coinscope
C
US
A B
C
GETDATA TX INV
HOW? INVBLOCKING
SLIDE 87
INVBLOCKING
A B
Coinscope
C
US
SLIDE 88
INVBLOCKING
A B
Coinscope
C
US
SLIDE 89
INV(txP) INV(txP) INV(txP)
INVBLOCKING
A B
Coinscope
C
US
SLIDE 90
INVBLOCKING
A B
Coinscope
C
US
SLIDE 91
INVBLOCKING
A B
Coinscope
C
US
txP txP txP
SLIDE 92
GETDATA (txP) GETDATA (txP) GETDATA (txP)
INVBLOCKING
A B
Coinscope
C
US
txP txP txP
SLIDE 93
INVBLOCKING
A B
Coinscope
C
US
txP txP txP
SLIDE 94
INVBLOCKING
A B
Coinscope
C
US
txP txP txP
We have a 2-min window where isolation and synchrony are not a problem!
SLIDE 95
∅
B’s Mempool
txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
∅
C’s Mempool
∅
A’s Mempool B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 96
∅
B’s Mempool
txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
txP
∅
C’s Mempool
∅
A’s Mempool B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 97
∅
B’s Mempool
txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
∅
C’s Mempool
∅
A’s Mempool B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 98
∅
B’s Mempool
txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
∅
C’s Mempool
txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 99
∅
B’s Mempool
txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
INV(txP )
∅
C’s Mempool
txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 100
∅
B’s Mempool
txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
∅
C’s Mempool
txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 101
∅
B’s Mempool
txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
∅
∅
C’s Mempool
txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 102
∅
B’s Mempool
txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
∅
C’s Mempool
txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 103
∅
B’s Mempool
txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
txF
∅
C’s Mempool
txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 104
∅
B’s Mempool
txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
∅
C’s Mempool
txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 105
∅
B’s Mempool
txF txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
∅
C’s Mempool
txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 106
∅
B’s Mempool
txF txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
txF
∅
C’s Mempool
txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 107
∅
B’s Mempool
txF txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
∅
C’s Mempool
txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 108
∅
B’s Mempool
txF txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
∅
C’s Mempool
txF txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 109
∅
B’s Mempool
txF txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
txF
∅
C’s Mempool
txF txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 110
∅
B’s Mempool
txF txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
txF
∅
C’s Mempool
txF txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 111
∅
B’s Mempool
txF txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
txF
∅
C’s Mempool
txF txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 112
∅
B’s Mempool
txF txP txP txP
SIMPLIFIED TXPROBE
A B
Coinscope
C
US
∅
C’s Mempool
txF txP
∅
A’s Mempool
txP
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 113
∅
B’s Mempool
txF txP txP txP
A B
Coinscope
C
US
∅
C’s Mempool
txF txP
∅
A’s Mempool
txP txM
SIMPLIFIED TXPROBE
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 114
∅
B’s Mempool
txF txP txP txP
A B
Coinscope
C
US
∅
C’s Mempool
txF txP
∅
A’s Mempool
txP
SIMPLIFIED TXPROBE
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 115
∅
B’s Mempool
txF txP txP txP
A B
Coinscope
C
US
∅
C’s Mempool
txF txP
∅
A’s Mempool
txP txM
SIMPLIFIED TXPROBE
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 116
∅
B’s Mempool
txF txP txP txP
A B
Coinscope
C
US
∅
C’s Mempool
txF txP
∅
A’s Mempool
txP txM txM
SIMPLIFIED TXPROBE
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 117
∅
B’s Mempool
txF txP txP txP
A B
Coinscope
C
US
∅
C’s Mempool
txF txP
∅
A’s Mempool
txP txM
SIMPLIFIED TXPROBE
B’s Orphanpool
∅
C’s Orphanpool
∅
SLIDE 118
∅
B’s Mempool
txF txP txP txP
A B
Coinscope
C
US
∅
C’s Mempool
txF txP
∅
A’s Mempool
txP txM
SIMPLIFIED TXPROBE
B’s Orphanpool
∅
C’s Orphanpool
∅
txM
SLIDE 119
∅
B’s Mempool
txF txP txP txP
A B
Coinscope
C
US
∅
C’s Mempool
txF txP
∅
A’s Mempool
txP txM
SIMPLIFIED TXPROBE
B’s Orphanpool
∅
C’s Orphanpool
∅
txM
SLIDE 120
- Choose a target node
- Create Parent, Marker and Flood transactions
- INVBLOCK the network
- Send Flood to all connected nodes but target
- Let Flood propagate
- Send Parent to target
- Send Marker to target
- Let Marker propagate
- Request marker back from all nodes but target
TXPROBE - PROTOCOL OVERVIEW
SLIDE 121
- Choose a target node
- Create Parent, Marker and Flood transactions
- INVBLOCK the network
- Send Flood to all connected nodes but target
- Let Flood propagate
- Send Parent to target
- Send Marker to target
- Let Marker propagate
- Request marker back from all nodes but target
TXPROBE - PROTOCOL OVERVIEW
For every node in the network
SLIDE 122
For a network like Bitcoin mainnet: nodes ≈ 10000 time ≈ 8.25 hours cost = 573210-764280 satoshi (5 sat/byte) ≈ $(20-30)
TXPROBE - COSTS ESTIMATION
SLIDE 123
We run 5 Bitcoin Core nodes as ground truth We define our precision / recall by checking how well can we infer the ground truth nodes connections Over 40 trials and with 95% confidence:
- Precision = 100%
- Recall = 93.86% - 95.45%
TXPROBE - DATA VALIDATION (TESTNET)
SLIDE 124
TXPROBE - TESTNET TOPOLOGY
precision = 100% recall = 97.40% size degree color Community unfolding
Higher community structure and modularity than random graph
SLIDE 125
FIXES
SLIDE 126
FIXES
SLIDE 127
FIXES
Truly randomised orphan transaction eviction
SLIDE 128
FIXES
SLIDE 129
FIXES
SLIDE 130
FIXES
Fixes INVBLOCKING
SLIDE 131
FIXES
SLIDE 132
FIXES
SLIDE 133
FIXES
Creates block-only relay connections
SLIDE 134
FIXES
SLIDE 135
FIXES
Bitcoin Core 0.18.0 Bitcoin Core 0.19.0 More info about the fixes: https://bitcoinops.org/en/newsletters/2019/09/18/ h/t David A. Harding (@hrdng)
SLIDE 136
sr_gi
QUESTIONS
#SFBW19
SLIDE 137
Testnet vs Mainnet INVBLOCKING (no-link) Efficiency / Orphan pool eviction
BONUS TRACK
SLIDE 138
- TxProbe is rather invasive: it empties the MapOrphanTransactions
pool of all nodes in the network every round
- We could not measure the implication that such behavior may have
had on the propagation of regular transactions
WHY TESTNET AND NO MAINNET?
SLIDE 139
INVBLOCKING V2
A B
Coinscope
C
US
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
SLIDE 140
INV(txF) INV(txP)
INVBLOCKING V2
A B
Coinscope
C
US
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
SLIDE 141
INVBLOCKING V2
A B
Coinscope
C
US
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
SLIDE 142
INVBLOCKING V2
A B
Coinscope
C
US
txP txF
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
SLIDE 143
GETDATA (txF) GETDATA (txP)
INVBLOCKING V2
A B
Coinscope
C
US
txP txF
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
SLIDE 144
INVBLOCKING V2
A B
Coinscope
C
US
txP txF
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
SLIDE 145
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
txF
SLIDE 146
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
SLIDE 147
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
∅
txF
SLIDE 148
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
txF
∅
txF
SLIDE 149
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
∅
txF
SLIDE 150
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
∅
txF
∅
txF
SLIDE 151
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
INV(txF)
∅
txF
∅
txF
SLIDE 152
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
∅
txF
∅
txF
SLIDE 153
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
∅
∅
txF
∅
txF
SLIDE 154
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
∅
txF
∅
txF
SLIDE 155
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
txP
∅
txF
∅
txF
SLIDE 156
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
∅
txF
∅
txF
SLIDE 157
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
∅
txF
∅
txF
∅
txP
SLIDE 158
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
txP
∅
txF
∅
txF
∅
txP
SLIDE 159
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
txP
∅
txF
∅
txF
∅
txP
SLIDE 160
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
txP
∅
txF
∅
txF
∅
txP
SLIDE 161
∅
A’s Mempool
∅
C’s Mempool
∅
B’s Mempool
txP txF
INVBLOCKING V2
A B
Coinscope
C
US
∅
txF
∅
txF
∅
txP
SLIDE 162
INVBLOCKING COMPARISON
A B
Coinscope
C
US
A B
Coinscope
C
US
∅
A’s Mempool
txP
∅
B’s Mempool
txF
∅
C’s Mempool
txF
∅
A’s Mempool
txP
∅
B’s Mempool
txF
∅
C’s Mempool
txF
SLIDE 163
Isolation
MAKE THIS WORK IN A REAL NETWORK
Synchrony Efficiency
A
txP
A
txP
B
txF
≈ O(n)
≈ O( n)
n = #nodes
SLIDE 164
ORPHANPOOL EVICTION (BEFORE v0.18)
while (mapOrphanTransactions.size() > nMaxOrphans) { // Evict a random orphan: uint256 randomhash = rng.rand256(); std::map<uint256, COrphanTx>::iterator it = mapOrphanTransactions.lower_bound(randomhash); if (it == mapOrphanTransactions.end()) it = mapOrphanTransactions.begin(); EraseOrphanTx(it->first); ++nEvicted; }
source: https://github.com/bitcoin/bitcoin/blob/273d025/src/net_processing.cpp#L783-L791
SLIDE 165
ORPHANPOOL EVICTION (BEFORE v0.18)
while (mapOrphanTransactions.size() > nMaxOrphans) { // Evict a random orphan: uint256 randomhash = rng.rand256(); std::map<uint256, COrphanTx>::iterator it = mapOrphanTransactions.lower_bound(randomhash); if (it == mapOrphanTransactions.end()) it = mapOrphanTransactions.begin(); EraseOrphanTx(it->first); ++nEvicted; }
source: https://github.com/bitcoin/bitcoin/blob/273d025/src/net_processing.cpp#L783-L791
- Pick a random 256-bit value R
- Get the orphan transaction (O) with hash closer to, but greater
than, R
- Evict O
- Repeat until mapOrphanTransaction is not full (default: 100)
- Double-spends are not checked for orphans
SLIDE 166
TXPROBE TRANSACTIONS OVERVIEW
Markeri
Flood
UTXO1
Parenti Markeri
UTXO0
Cleanser
Markeri
Squatteri
SLIDE 167
MAKE ROOM IN THE ORPHANPOOL
Cleanser
Markeri
Squatteri
- Create the cleanser (regular transaction) and 100 squatters (double-
spends between each other)
- Every squatter is created in a POW-ish way (e.g. re-sign until its hash
falls bellow a certain threshold)
- All squatters are sent to the flood set nodes to replace any existing
- rphan.
- Finally, the cleanser is sent to empty the orphanpool
SLIDE 168
sr_gi
QUESTIONS
#SFBW19