Program Analysis for Web3
Dr. Saurabh Joshi
Principal Researcher
Supra Research
What is a blockchain?
Execution trace of a deterministic state machine
What is a blockchain?
s_1
s_2
s_3
s_4
s_4
s_5
t_1
t_3
t_5
t_2
t_4
What is a blockchain?
Programs are state transformers
x=x+1x = 0
x = 1
x=x+1x = 1
x = 2
x=x+1x = 255
x = ?
Need for determinism
Without deterministic operational semantics, given the same sequence of transactions, output state may be different when executed on a different hardware, or on the same hardware at a different time
What is a blockchain?
Execution trace of a deterministic state machine agreed upon by a network of distributed, possibly heterogenous and decentralized computers
Determinism
- Floating point arithmetic
- Signed integer arithmetic
- Concurrency
- Interaction with external environment
- ..and many more
Termination
- All transactions are executed with a pre-determined upperbound on gas (resources) consumed
gas : Instr \rightarrow \mathbb{N}
- Can we perform a static analysis on a code that given a fixed gas cost table, a given code would terminate within a specified gas limit? If the code is parametric, for what value of n it can terminate within the gas limit?
- Challenges with respect to native functions.
Gas Optimization
- Reduce the amount of gas used while retaining the same input/output behaviour
minimize \sum gas(Instr_i)
- Gas for storage and gas for computation makes gas optimization a bit tricky.
- Would caching be cheaper as compared to computing a value every time?
Gas Estimation
- Provide the number of gas units that would be consumed by the transaction
- Undecidability
- Input context is not known until execution
- Provide an estimate for the number of gas units that would be consumed by the transaction
- Accurate estimation needed both for users and for the network
Improve execution
Deriving access specification
Part 1: Given a program P(s), derive abstract access specifications in terms of ReadSet and WriteSet
Part 2: Given a program P(s), and input state s, quickly refine the abstract specification to minimize ReadSet and WriteSet as much as possible
Compilers and Virtual Machines
- From high level programming languages to blockchain VMs (e.g., Solidity -> EVM, Move -> Move VM, Rust/C++ -> Solana VM)
- Enabling smart contracts written in one language to work on another VM (e.g., Fractal, A transpiler that makes Solidity work on Move VM)
- Enabling smart contracts written in one language to work on near native assembly (e.g., Rust/C++ -> Wasm)
Safety and security of smart contracts
- Static analysis tools for vulnerability detection (e.g., Slither, Mythrill, Smartify)
- Model-checkers (e.g., ESBMC-Solidity, Zeus)
- Provers (e.g., Move Prover)
- https://medium.com/immunefi/the-top-10-most-common-vulnerabilities-in-web3-bf7a921d489f
Example Vulnerability
function multiClaim(uint256[] memory _tickets) external {
uint256 totalReward = 0;
for (uint i = 0; i < _tickets.length; i++) {
require (msg.sender == lotteryNFT.ownerOf(_tickets[i]), “not from owner”);
require (!lotteryNFT.getClaimStatus(_tickets[i]), “claimed”);
uint256 reward = getRewardView(_tickets[i]);
if(reward>0) {
totalReward = reward.add(totalReward);
}
}
lotteryNFT.multiClaimReward(_tickets);
if(totalReward>0) {
cake.safeTransfer(address(msg.sender), totalReward);
}
emit MultiClaim(msg.sender, totalReward);
}“If we want to be serious about quality, it is time to get tired of finding bugs and start preventing their happening in the first place.”— Alan Page
Program Analysis for Web3
By Saurabh Joshi
