CS251 Fall 2021

(https://cs251.stanford.edu)

Cryptocurrencies and
Blockchain Technologies
Dan Boneh Benedikt Bünz
Stanford University

[videos on canvas, discussions on edstem, homework on gradescope]

 What is a blockchain?

Abstract answer: a blockchain provides

• coordination between many parties,
• when there is no single trusted party

if trusted party exists ⇒ no need for a blockchain

[financial systems: often no trusted party]

 What is all the excitement about?

(1) Basic application: a digital currency (stored value)

• Current largest: Bitcoin (2009), Ethereum (2015)
• Global: accessible to anyone with an Internet connection
 What is all the excitement about?

(2) Beyond stored value: decentralized applications (DAPPs)

• DeFi: financial instruments managed by public programs
• examples: stablecoins, lending, exchanges, ….
• Asset management (NFTs): art, game assets, domain names.
• Decentralized organizations (DAOs): (decentralized governance)
• DAOs for investment, for donations, for collecting art, etc.

(3) New programming model: writing decentralized programs

Assets managed by DAPPs

Sep. 2021
Transaction volume

24h volume Sep. 2021

 Central Bank Digital Currency (CBDC)

e ta il CB DC
r k i n g o n r
c ti vel y w o 9]
ra l b a nks a vey Jan . 2 0 1
30 c e nt [B I S s u r
What is a blockchain?

user facing tools (cloud servers)

applications (DAPPs, smart contracts)

compute layer (blockchain computer)

consensus layer
 Consensus layer (informal)

achieved by replication
A public append-only data structure:
• Persistence: once added, data can never be removed*
• Safety: all honest participants have the same data**
• Liveness: honest participants can add new transactions
• Open(?): anyone can add data (no authentication)

consensus layer
 How are blocks added to chain?
blockchain

I am the
signed leader
2 ETH
skA verify
verify block
block
skB

skC
 How are blocks added to chain?
blockchain
…
2 ETH I am the
leader
skA

skB

2 ETH
skC
 Why is consensus a hard problem?
Tx1, Tx2, Tx3, Tx4 Tx1, Tx2, Tx3, Tx4

Tx1 Tx3

Tx2 Tx4

Tx1, Tx2, Tx3, Tx4 Tx1, Tx2, Tx3, Tx4

 Why is consensus a hard problem?
Tx1, Tx2, Tx3, Tx4 Tx1, Tx2, Tx3, Tx4

Tx1
Problems:
∆ Tx3
• Network delay
• Network partition
Tx2 Tx4
∆

Tx1, Tx2, Tx3, Tx4 Tx1, Tx2, Tx3, Tx4

 Why is consensus a hard problem?
Tx1, Tx2, Tx4

Tx1
Problems:
• crash
• malice
Tx2 Tx4

Tx1, Tx2, Tx4 Tx1, Tx2, Tx4

 The blockchain computer
DAPP logic is encoded in a program that runs on blockchain
• Rules are enforced by a public program (public source code)
⇒ transparency: no single trusted 3rd party

• The DAPP program is executed by parties who create new blocks

⇒ public verifiability: everyone can verify state transitions

compute layer
consensus layer
Decentralized applications (DAPPS)

Run on
blockchain
computer

applications (DAPPs, smart contracts)

blockchain computer

consensus layer
 Common DAPP architecture
Top layer: user facing servers

end user

DAPP DAPP DAPP

on-chain
state blockchain computer
consensus layer
[source: the Block Genesis]
lots of experiments …
locked
 This course

Cryptography Distributed systems

Economics
 Course organization
1. The starting point: Bitcoin mechanics
2. Consensus protocols
3. Ethereum and decentralized applications
4. Economics of decentralized applications
5. Scaling the blockchain: 10K Tx/sec
6. Private transactions on a public blockchain
(SNARKs and zero knowledge proofs)
7. Interoperability among chains: bridges and wrapped coins
 Course organization

cs251.stanford.edu

• Three homework problems, four projects, final exam(?)

• Optional weekly sections on Friday

Please tell us how we can improve …

Don’t wait until the end of the quarter
Let’s get started …
 Cryptography Background

(1) cryptographic hash functions

An efficiently computable function

where
32 bytes
256
megabytes hash value 𝑇 = {0,1 }
 Collision resistance

Def: a collision for is pair s.t.

implies that many collisions exist

Def: a function is collision resistant if it is “hard” to find even a

single collision for (we say is a CRHF)

Example: SHA256: { : len()<264 bytes} ⇾ {0,1}256

details in CS255
 Application: committing to data on a
blockchain
Alice has a large file . She posts (32 bytes)
Bob reads . Later he learns s.t.

is a CRHF ⇒ Bob is convinced that

(otherwise, and are a collision for )

We say that is a binding commitment to

(note: not hiding, may leak information about )

 Committing to a list (of transactions)

Alice has 32 bytes

Goal:
- Alice posts a short binding commitment to ,
- Bob reads . Given can check that
Bob runs ⇾ accept/reject

security: adv. cannot find s.t. and

where
 Merkle tree (Merkle 1989)

commitment h Goal:
• commit to list S of size n
• Later prove

Merkle tree
commitment

𝑚1 𝑚2 𝑚3 𝑚4 𝑚5 𝑚6 𝑚7 𝑚8
list of values S
 Merkle tree (Merkle 1989)

commitment h Goal:
• commit to list S of size n
𝑦5 H 𝑦6 • Later prove

𝑦1 H 𝑦 𝑦3 H 𝑦 4
2 To prove ,
H H H H proof
𝑚1 𝑚2 𝑚3 𝑚4 𝑚5 𝑚6 𝑚7 𝑚8 length of proof:

list of values S
 Merkle tree (Merkle 1989)

commitment h To prove ,
proof
𝑦5 H 𝑦6
𝑦1 H 𝑦 𝑦3 H 𝑦 4 Bob does:
2

H H H H

𝑚1 𝑚2 𝑚3 𝑚4 𝑚5 𝑚6 𝑚7 𝑚8
accept if
list of values S
 Merkle tree (Merkle 1989)

Thm: H CRHF ⇒ adv. cannot find s.t. and

where

(to prove, prove the contra-positive)

How is this useful? Super useful. Example

• When writing a block of transactions to the blockchain,
suffices to write commit() to chain. Keep chain small.
• Later, can prove contents of every Tx.
 Abstract block chain
blockchain
block header block header block header
Merkle other Merkle other Merkle other
⊥ hash hash
root data root data root data

Merkle Merkle Merkle

tree tree tree

Tx1 Tx2 … Txn Tx1 Tx2 … Txn Tx1 Tx2 … Txn

Merkle proofs are used to prove that a Tx is “on the block chain”
 Another application: proof of work

Goal: computational problem that

• takes time to solve, but (D is called the difficulty)
• solution takes time O(1) to verify

How? e.g.

• puzzle: input , output s.t.

• verify(): accept if
 Another application: proof of work

Thm: if H is a “random function” then the best algorithm

requires evaluations of in expectation.

Note: this is a parallel algorithm

⇒ the more machines I have, the faster I solve the puzzle.

Proof of work is used in some consensus protocols (e.g., Bitcoin)

Bitcoin uses
Cryptography background:
Digital Signatures
How to authorize a transaction
 Signatures
Physical signatures: bind transaction to author

Bob agrees to pay Alice 1$

Bob agrees to pay Alice 100$

Problem in the digital world:

anyone can copy Bob’s signature from one doc to another
 Digital signatures
Solution: make signature depend on document
Signer
Verifier
‘accept’
Bob agrees to pay Alice 1$
verifier or
‘reject’
signature

signing
secret signing
algorithm public verification
key (sk) key (pk)
 Digital signatures: syntax
Def: a signature scheme is a triple of algorithms:
• Gen(): outputs a key pair (pk, sk)
• Sign(sk, msg) outputs sig. σ
• Verify(pk, msg, σ) outputs ‘accept’ or ‘reject’

Secure signatures: (informal)

Adversary who sees signatures on many messages of his choice,
cannot forge a signature on a new message.
 Families of signature schemes
1. RSA signatures (old … not used in blockchains):
• long sigs and public keys (≥256 bytes), fast to verify

2. Discrete-log signatures: Schnorr and ECDSA (Bitcoin, Ethereum)

• short sigs (48 or 64 bytes) and public key (32 bytes)

3. BLS signatures: 48 bytes, aggregatable, easy threshold

(Ethereum 2.0, Chia, Dfinity)

4. Post-quantum signatures: long (≥768 bytes)

details in CS255
 Signatures on the blockchain
Signatures are used everywhere: verify
• ensure Tx authorization, Tx
• governance votes, verify
verify Tx
• consensus protocol votes. Tx

sk1 data signatures

sk2 data signatures

 END OF LECTURE

Next lecture: the Bitcoin blockchain