Bobtail: Improved Blockchain Security With Low-Variance Mining
GEORGE BISSIAS BRIAN LEVINE
UNIVERSITY OF MASSACHUSETTS AMHERST
Bobtail: Improved Blockchain Security With Low-Variance Mining - - PowerPoint PPT Presentation
Bobtail: Improved Blockchain Security With Low-Variance Mining GEORGE BISSIAS BRIAN LEVINE UNIVERSITY OF MASSACHUSETTS AMHERST Incentive for Compressed Review of Blockchains block proposal We focus on public / open blockchains BLOCK
GEORGE BISSIAS BRIAN LEVINE
UNIVERSITY OF MASSACHUSETTS AMHERST
Compressed Review of Blockchains
BLOCK ENTRY 1 ENTRY 2 . . . ENTRY N
▸ We focus on public / open blockchains
that use proof-of-work (PoW)
▸ Decentralized and distributed ledgers ▸ Ledger comprises set of transactions ▸ Financial, logistical, legal, … ▸ PoW: not the only approach, but most
popular and relatively easy to analyze
B 1 BTC A
Incentive for block proposal
BOBTAIL
Proof-of-Work Mining Basics
▸ Miners repeatedly hash block header ▸ Hashes are within ▸ A block is mined when hash falls
below
▸ Block time is function of hash rate
(seconds)
▸ Convention is to extend longest chain
[0,S] t T h
HASH BLOCK HEADER BLOCK
t v
T
h
BLOCK Transactions
BOBTAIL
Mining is a Lottery
▸ Miners “draw” numbers
until they cross threshold 5
BOBTAIL
Mining is a Lottery
▸ Miners “draw” numbers
until they cross threshold 5
▸ Each draw “costs” a hash
17
BOBTAIL
Mining is a Lottery
▸ Miners “draw” numbers
until they cross threshold 5
▸ Each draw “costs” a hash
17 42
BOBTAIL
Mining is a Lottery
▸ Miners “draw” numbers
until they cross threshold 5
▸ Each draw “costs” a hash ▸ First to cross threshold wins ▸ Winner receives a reward
and proposes a block 17 42
. . .
3
BOBTAIL
Mining is a Lottery
▸ Miners “draw” numbers
until they cross threshold 5
▸ Each draw “costs” a hash ▸ First to cross threshold wins ▸ Winner receives a reward
and proposes a block
▸ Game repeats
. . . . . . . . .
BOBTAIL
Mining statistics
▸ Time to draw below threshold
is approximately
▸ 20% miner expects to take 4
times as long to mine a block as
Expon (
T q )
5 4 T
5T
5 4 T
(Individual) (Others)
BOBTAIL
Double-spending Attack
▸ Alice trades car for 1 BTC ▸ Transaction appears in block 1 ▸ Assumes majority are mining
chain
▸ Alice knows about law of large
numbers
▸ Goods are released only once
payment has “confirmations”
z
1 …
Alice (merchant) Bob (attacker)
Transaction
B 1 BTC A
z
BOBTAIL
Double-spending Attack
▸ Bob steals goods if red chain
grows longer than blue
▸ Relies on high variance of the
exponential distribution
▸ Goods worth more than cost
1 … 1
Alice (merchant) Bob (attacker)
Transaction
B 1 BTC A
2
Transaction
B 1 BTC B′
z
BOBTAIL
Attack Success Probability
▸ Attacker needs to get ahead
by at least one block sometime after the first blocks
▸ Even a 20% miner has 5%
chance of winning after 6 blocks
z
k = 1 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% 80.0% 90.0%
. of succcessful doublespend
1 3 5 7 9 11
Embargo Period z
attacker mining power
0.1 0.2 0.3 0.4 0.45
BOBTAIL
Bobtail Protocol Details
▸ Assemble a block containing transactions ▸ Hash header as usual to generate
“proofs”
▸ Disseminate proofs that are “low enough”
to neighbors
▸ Maintain queue of lowest proofs ▸ Assemble proofs whose mean is below ▸ Each proof miner receives reward
k k t
tk
1 k ∑
i
pi
T
PROOF BLOCK
h
BLOCK
. . .
PROOF BLOCK
h
p1 p2 p5 p3 p4 p1 p2 p3 p4 p5
BOBTAIL
New Lottery: Bobtail
▸ Miners draw numbers until
the average of any 2 cross threshold 5
BOBTAIL
New Lottery: Bobtail
9
▸ Miners draw numbers until
the average of any 2 cross threshold 5
▸ Each draw still “costs” a hash
BOBTAIL
New Lottery: Bobtail
3 9
▸ Miners draw numbers until
the average of any 2 cross threshold 5
▸ Each draw still “costs” a hash
BOBTAIL
New Lottery: Bobtail
3 9 12
. . .
▸ Miners draw numbers until
the average of any 2 cross threshold 5
▸ Each draw still “costs” a hash
BOBTAIL
New Lottery: Bobtail
▸ Miners draw numbers until
the average of any 2 cross threshold 5
▸ Each draw still “costs” a hash ▸ First 2 to cross threshold win ▸ Winners receive a reward
and lowest proposes a block 3 9 12 6
. . .
3
BOBTAIL
Impact on Doublespend Attack Efficacy
▸ Status quo (Bitcoin) ▸ 20% attacker succeeds
approximately 5% of the time after 6 confirmations
▸ Bobtail with k=20 ▸ 20% attacker succeeds less than
1% of the time with just 2 confirmations
attacker mining power
0.1 0.2 0.3 0.4 0.45
k = 1 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% 80.0% 90.0%
. of succcessful doublespend
1 3 5 7 9 11
Embargo Period z
1 3 5 7 9 11 k = 20
BOBTAIL
Relative Statistics
▸ Mining time with Bobtail for fixed target : ▸ Expected value increases by ▸ Variance increases by ▸ When expected times are aligned: ▸ ▸ Relative variance
t k + 1 2 (k + 1)(2k + 1) 6k tk = k + 1 2 t O(1/k)
5 10 15 20 25 30 35 40 45 50 55 60 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 5 10 15 20 25 30 35 40
Ethereum (seconds per block) Bitcoin (minutes per block) CDF k
1 10 20 40 5
BOBTAIL
What is the Cost?
▸ Size of meta data increases by
160B
k ⋅
BOBTAIL
What is the Cost?
▸ Size of meta data increases by
160B
▸ Increased network overhead ▸ Mitigated by not sending proofs in the “tail” ▸ Graphene can be used to reduce redundancy
k ⋅
Gamma shape k Don’t send Send
BOBTAIL
What is the Cost?
▸ Size of meta data increases by
160B
▸ Increased network overhead ▸ New attacks must be considered ▸ Proof withholding ▸ Denial-of-Service (DoS)
k ⋅
Summary
▸ Mining process is akin to a lottery ▸ We can skew statistics in favor of honest majority ▸ This greatly mitigates fundamental attacks ▸ Doublespend susceptibility reduced by orders of magnitude ▸ Primary cost is increased network and block overhead