multiversx-security-audit（多维安全审计）

MultiversX Security Audit Methodology

A sequential workflow for auditing MultiversX smart contracts, from initial reconnaissance through automated scanning.

When to Use

• Starting a new security audit engagement
• Performing security code reviews
• Setting up automated vulnerability scanning
• Mapping attack surface for penetration testing
• Training new security reviewers

Phase 1: Context Building

Rapidly build a comprehensive mental model of the codebase before diving into vulnerability hunting.

1.1 Reconnaissance Checklist

• Core logic: #[multiversx_sc::contract], #[payable], value-moving functions, access control patterns
• Externalities: cross-contract calls, hardcoded addresses (sc: literals), oracle dependencies, bridge contracts
• Documentation: README.md, specs/, sc-config.toml, multiversx.yaml, snippets.sh, scenarios/ tests

1.2 System Mapping

Roles and Permissions

Asset Inventory

State Analysis

Document all storage mappers and their purposes:

// Example state inventory
#[storage_mapper("owner")]           // SingleValueMapper - access control
#[storage_mapper("balances")]        // MapMapper - user funds (CRITICAL)
#[storage_mapper("whitelist")]       // SetMapper - privileged users

1.3 Entry Point Enumeration (Initial)

List all #[endpoint] functions with their risk classification:

HIGH RISK:
- deposit() - #[payable("*")] - accepts value
- withdraw() - moves funds out
- upgrade() - can change contract logic

MEDIUM RISK:
- setConfig() - owner only, changes behavior
- addWhitelist() - expands permissions

LOW RISK:
- getBalance() - #[view] - read only

1.4 Threat Modeling (Initial)

Asset at Risk Analysis

• Direct Loss: What funds can be stolen if the contract fails?
• Indirect Loss: What downstream systems depend on this contract?
• Reputation Loss: What non-financial damage could occur?

Attacker Profiles

1.4 Environment Check

Dependency Audit

• Framework Version: Check Cargo.toml for multiversx-sc version
• Known Vulnerabilities: Compare against security advisories
• Deprecated APIs: Look for usage of deprecated functions

Test Suite Assessment

• Coverage: Does scenarios/ exist with comprehensive tests?
• Edge Cases: Are failure paths tested?
• Freshness: Run sc-meta test-gen to verify tests match current code

1.5 Context Building Output

After completing context building, document:

1. System Overview: One-paragraph summary of what the contract does
2. Trust Boundaries: Who trusts whom, what assumptions exist
3. Critical Paths: The most security-sensitive code paths
4. Initial Concerns: Preliminary list of areas requiring deep review
5. Questions for Team: Clarifications needed from developers

Context Building Checklist

• All entry points identified and classified
• Storage layout documented
• External dependencies mapped
• Role/permission model understood
• Test coverage assessed
• Framework version noted
• Initial threat model drafted

Phase 2: Entry Point Analysis

Identify the complete attack surface by enumerating all public interaction points and classifying their risk levels.

2.1 Entry Point Identification

MultiversX Macros That Expose Functions

Scanning Commands

# Find all endpoints
grep -n "#\[endpoint" src/*.rs

# Find all payable endpoints
grep -n "#\[payable" src/*.rs

# Find all views
grep -n "#\[view" src/*.rs

# Find callbacks
grep -n "#\[callback" src/*.rs

# Find init and upgrade
grep -n "#\[init\]\|#\[upgrade\]" src/*.rs

2.2 Risk Classification

Category A: Payable Endpoints (Critical Risk)

Functions receiving value require the most scrutiny.

#[payable]
#[endpoint]
fn deposit(&self) {
    // MUST CHECK:
    // 1. Token identifier validation
    // 2. Amount > 0 validation
    // 3. Correct handling of multi-token transfers
    // 4. State updates before external calls

    let payment = self.call_value().single();
    require!(
        payment.token_identifier == self.accepted_token().get(),
        "Wrong token"
    );

    // Process deposit...
}

Checklist for Payable Endpoints:

• Token ID validated against expected token(s)
• Amount checked for minimum/maximum bounds
• Multi-transfer handling if all() used
• Nonce validation for NFT/SFT
• Reentrancy protection (Checks-Effects-Interactions)

Category B: Non-Payable State-Changing Endpoints (High Risk)

#[endpoint]
fn update_config(&self, new_value: BigUint) {
    // MUST CHECK:
    // 1. Who can call this? (access control)
    // 2. Input validation
    // 3. State transition validity

    self.require_caller_is_admin();
    require!(new_value > 0, "Invalid value");
    self.config().set(new_value);
}

Checklist for State-Changing Endpoints:

• Access control implemented and correct
• Input validation for all parameters
• State transitions are valid
• Events emitted for important changes
• No DoS vectors (unbounded loops, etc.)

Category C: View Functions (Low Risk)

#[view(getBalance)]
fn get_balance(&self, user: ManagedAddress) -> BigUint {
    // SHOULD CHECK:
    // 1. Does it actually modify state? (interior mutability)
    // 2. Does it leak sensitive information?
    // 3. Is the calculation expensive (DoS via gas)?

    self.balances(&user).get()
}

Checklist for View Functions:

• No state modification (verify no storage writes)
• No sensitive data exposure
• Bounded computation (no unbounded loops)
• Block info usage appropriate (get_block_timestamp_millis() / get_block_timestamp_seconds() may differ off-chain)

Category D: Init and Upgrade (Critical Risk)

#[init]
fn init(&self, admin: ManagedAddress) {
    // MUST CHECK:
    // 1. All required state initialized
    // 2. No way to re-initialize
    // 3. Admin/owner properly set

    self.admin().set(admin);
}

#[upgrade]
fn upgrade(&self) {
    // MUST CHECK:
    // 1. New storage mappers initialized
    // 2. Storage layout compatibility
    // 3. Migration logic correct
}

Category E: Callbacks (High Risk)

#[callback]
fn transfer_callback(
    &self,
    #[call_result] result: ManagedAsyncCallResult<()>
) {
    // MUST CHECK:
    // 1. Error handling (don't assume success)
    // 2. State reversion on failure
    // 3. Correct identification of original call

    match result {
        ManagedAsyncCallResult::Ok(_) => {
            // Success path
        },
        ManagedAsyncCallResult::Err(_) => {
            // CRITICAL: Must handle failure!
            // Revert any state changes from original call
        }
    }
}

2.3 Analysis Workflow

Step 1: List All Entry Points

| Endpoint | Type | Payable | Access | Storage Touched | Risk |
|----------|------|---------|--------|-----------------|------|
| deposit | endpoint | * | Public | balances | Critical |
| withdraw | endpoint | No | Public | balances | Critical |
| setAdmin | endpoint | No | Owner | admin | High |
| getBalance | view | No | Public | balances (read) | Low |
| init | init | No | Deploy | admin, config | Medium |

Step 2: Tag Access Control

// Public - anyone can call
#[endpoint]
fn public_function(&self) { }

// Owner only - blockchain owner
#[only_owner]
#[endpoint]
fn owner_function(&self) { }

// Admin only - custom access control
#[endpoint]
fn admin_function(&self) {
    self.require_caller_is_admin();
}

// Whitelisted - address in set
#[endpoint]
fn whitelist_function(&self) {
    let caller = self.blockchain().get_caller();
    require!(self.whitelist().contains(&caller), "Not whitelisted");
}

Step 3: Tag Value Handling

Step 4: Graph Data Flow

deposit() ──writes──▶ balances
          ──writes──▶ total_deposited
          ──reads───▶ accepted_token

withdraw() ──reads/writes──▶ balances
           ──reads────────▶ withdrawal_fee

getBalance() ──reads──▶ balances

2.4 Specific Attack Vectors

Privilege Escalation

// VULNERABLE: Missing access control
#[endpoint]
fn set_admin(&self, new_admin: ManagedAddress) {
    self.admin().set(new_admin);  // Anyone can become admin!
}

// CORRECT: Protected
#[only_owner]
#[endpoint]
fn set_admin(&self, new_admin: ManagedAddress) {
    self.admin().set(new_admin);
}

DoS via Unbounded Growth

// VULNERABLE: Public endpoint adds to unbounded set
#[endpoint]
fn register(&self) {
    let caller = self.blockchain().get_caller();
    self.participants().insert(caller);  // Grows forever!
}

Missing Payment Validation

Bad:

// DON'T: Accept any token without validation — attacker sends worthless tokens
#[payable]
#[endpoint]
fn stake(&self) {
    let payment = self.call_value().single();
    self.staked().update(|s| *s += payment.amount.as_big_uint()); // Fake tokens accepted!
}

Good:

// DO: Validate token identity and optionally enforce minimum amount
#[payable]
#[endpoint]
fn stake(&self) {
    let payment = self.call_value().single();
    require!(
        payment.token_identifier == self.staking_token().get(),
        "Wrong token"
    );
    self.staked().update(|s| *s += payment.amount.as_big_uint());
}

Callback State Assumptions

// VULNERABLE: Assumes success
#[callback]
fn on_transfer_complete(&self) {
    // This runs even if transfer FAILED!
    self.transfer_count().update(|c| *c += 1);
}

2.5 Entry Point Analysis Output Template

# Entry Point Analysis: [Contract Name]

## Summary
- Total Endpoints: X
- Payable Endpoints: Y (Critical)
- State-Changing: Z (High)
- Views: W (Low)

## Access Control Matrix

| Endpoint | Public | Owner | Admin | Whitelist |
|----------|--------|-------|-------|-----------|
| deposit | Yes | - | - | - |
| setAdmin | - | Yes | - | - |

## Recommended Focus Areas
1. [Highest priority endpoint and why]
2. [Second priority]
3. [Third priority]

Phase 3: Static Analysis Patterns

Manual and automated static analysis for finding vulnerabilities in MultiversX Rust and Go code.

3.1 Rust Smart Contracts (multiversx-sc)

Critical Grep Patterns

Unsafe Code

# Unsafe blocks - valid only for FFI or specific optimizations
grep -rn "unsafe" src/

Risk: Memory corruption, undefined behavior Action: Require justification for each unsafe block

Panic Inducers

# Direct unwrap - can panic
grep -rn "\.unwrap()" src/

# Expect - also panics
grep -rn "\.expect(" src/

# Index access - can panic on out of bounds
grep -rn "\[.*\]" src/ | grep -v "storage_mapper"

Risk: Contract halts, potential DoS Action: Replace with unwrap_or_else(|| sc_panic!(...)) or proper error handling

Floating Point Arithmetic

grep -rn "f32" src/
grep -rn "f64" src/
grep -rn "as f32\|as f64" src/

Risk: Non-deterministic behavior, consensus failure Action: Use BigUint/BigInt for all calculations

Unchecked Arithmetic

grep -rn "[^_a-zA-Z]\+ [^_a-zA-Z]" src/  # Addition
grep -rn "[^_a-zA-Z]\- [^_a-zA-Z]" src/  # Subtraction
grep -rn "[^_a-zA-Z]\* [^_a-zA-Z]" src/  # Multiplication

Risk: Integer overflow/underflow Action: Use BigUint or checked arithmetic for all financial calculations

Map Iteration (DoS Risk)

grep -rn "\.iter()" src/
grep -rn "for.*in.*\.iter()" src/
grep -rn "\.collect()" src/

Risk: Gas exhaustion DoS Action: Add pagination or bounds checking

Logical Pattern Analysis (Manual Review)

Token ID Validation

grep -rn "call_value()" src/
grep -rn "\.single()" src/
grep -rn "\.all()" src/
grep -rn "\.array()" src/
grep -rn "\.single_optional()" src/
# Legacy patterns (may still be in older code)
grep -rn "all_esdt_transfers" src/
grep -rn "single_esdt" src/

For each occurrence, verify:

• Token ID checked against expected value (using TokenId comparison)
• Token nonce validated (for NFT/SFT)
• Amount validated (within bounds — non-zero is guaranteed by NonZeroBigUint in Payment)

// VULNERABLE
#[payable]
fn deposit(&self) {
    let payment = self.call_value().single();
    self.balances().update(|b| *b += payment.amount.as_big_uint());
    // No token ID check! Accepts any token
}

// SECURE
#[payable]
fn deposit(&self) {
    let payment = self.call_value().single();
    require!(
        payment.token_identifier == self.accepted_token().get(),
        "Wrong token"
    );
    // Note: amount > 0 check is no longer needed — Payment.amount is NonZeroBigUint
    self.balances().update(|b| *b += payment.amount.as_big_uint());
}

Callback State Assumptions

grep -rn "#\[callback\]" src/

For each callback, verify:

• Does NOT assume async call succeeded
• Handles error case explicitly
• Reverts state changes on failure if needed

Access Control

grep -rn "#\[endpoint\]" src/
grep -rn "#\[only_owner\]" src/

For each endpoint, verify:

• Appropriate access control applied
• Sensitive operations restricted
• Admin functions documented

Reentrancy (CEI Pattern)

grep -rn "\.send()\." src/
grep -rn "\.tx()" src/
grep -rn "async_call" src/

Verify Checks-Effects-Interactions pattern:

• All checks (require!) before state changes
• State changes before external calls
• No state changes after external calls in same function

3.2 Go Protocol Code (mx-chain-go)

Concurrency Issues

Goroutine Loop Variable Capture

grep -rn "go func" *.go

// VULNERABLE
for _, item := range items {
    go func() {
        process(item)  // item may have changed!
    }()
}

// SECURE
for _, item := range items {
    item := item  // Create local copy
    go func() {
        process(item)
    }()
}

Map Race Conditions

grep -rn "map\[" *.go | grep -v "sync.Map"

Determinism Issues

Map Iteration Order

grep -rn "for.*range.*map" *.go

Map iteration in Go is random. Never use for generating hashes, creating consensus data, or any deterministic output.

Time Functions

grep -rn "time.Now()" *.go

time.Now() is forbidden in block processing — use block.Header.TimeStamp instead.

3.3 Analysis Checklist

Smart Contract Review Checklist

Access Control

• All endpoints have appropriate access restrictions
• Owner/admin functions use #[only_owner] or explicit checks
• No privilege escalation paths

Payment Handling

• Token IDs validated in all #[payable] endpoints
• Amounts validated (non-zero, bounds)
• NFT nonces validated where applicable

Arithmetic

• No raw arithmetic on u64/i64 with external inputs
• BigUint used for financial calculations
• No floating point

State Management

• Checks-Effects-Interactions pattern followed
• Callbacks handle failure cases
• Storage layout upgrade-safe

Gas & DoS

• No unbounded iterations
• Storage growth is bounded
• Pagination for large data sets

Error Handling

• No unwrap() without justification
• Meaningful error messages
• Consistent error handling patterns

Protocol Review Checklist

Concurrency

• All shared state properly synchronized
• No goroutine loop variable capture bugs
• Channel usage is correct

Determinism

• No map iteration for consensus data
• No time.Now() in block processing
• No random number generation without deterministic seed

Memory Safety

• Bounds checking on slices
• No nil pointer dereferences
• Proper error handling

3.4 Vulnerability Categories Quick Reference

Phase 4: Automated Scanning with Semgrep

Create custom Semgrep rules to automatically detect MultiversX-specific security patterns and best practice violations.

4.1 Semgrep Basics for Rust

Rule Structure

rules:
  - id: rule-identifier
    languages: [rust]
    message: "Description of the issue and why it matters"
    severity: ERROR  # ERROR, WARNING, INFO
    patterns:
      - pattern: <code pattern to match>
    metadata:
      category: security
      technology:
        - multiversx

Pattern Syntax

4.2 Common MultiversX Rules

Unsafe Arithmetic Detection

rules:
  - id: mvx-unsafe-addition
    languages: [rust]
    message: "Potential arithmetic overflow. Use BigUint or checked arithmetic for financial calculations."
    severity: ERROR
    patterns:
      - pattern: $X + $Y
      - pattern-not: $X.checked_add($Y)
      - pattern-not: BigUint::from($X) + BigUint::from($Y)
    paths:
      include:
        - "*/src/*.rs"
    metadata:
      category: security
      subcategory: arithmetic
      cwe: "CWE-190: Integer Overflow"

  - id: mvx-unsafe-multiplication
    languages: [rust]
    message: "Potential multiplication overflow. Use BigUint or checked_mul."
    severity: ERROR
    patterns:
      - pattern: $X * $Y
      - pattern-not: $X.checked_mul($Y)
      - pattern-not: BigUint::from($X) * BigUint::from($Y)

Floating Point Detection

rules:
  - id: mvx-float-forbidden
    languages: [rust]
    message: "Floating point arithmetic is non-deterministic and forbidden in smart contracts."
    severity: ERROR
    pattern-either:
      - pattern: "let $X: f32 = ..."
      - pattern: "let $X: f64 = ..."
      - pattern: "$X as f32"
      - pattern: "$X as f64"
    metadata:
      category: security
      subcategory: determinism

Payable Endpoint Without Value Check

rules:
  - id: mvx-payable-no-check
    languages: [rust]
    message: "Payable endpoint does not check payment value. Verify token ID and amount."
    severity: WARNING
    patterns:
      - pattern: |
          #[payable]
          #[endpoint]
          fn $FUNC(&self, $...PARAMS) {
              $...BODY
          }
      - pattern-not: |
          #[payable]
          #[endpoint]
          fn $FUNC(&self, $...PARAMS) {
              <... self.call_value() ...>
          }
    metadata:
      category: security
      subcategory: input-validation

Unsafe Unwrap Usage

rules:
  - id: mvx-unsafe-unwrap
    languages: [rust]
    message: "unwrap() can panic. Use unwrap_or_else with sc_panic! or proper error handling."
    severity: ERROR
    patterns:
      - pattern: $EXPR.unwrap()
      - pattern-not-inside: |
          #[test]
          fn $FUNC() { ... }
    fix: "$EXPR.unwrap_or_else(|| sc_panic!(\"Error message\"))"
    metadata:
      category: security
      subcategory: error-handling

Missing Owner Check

rules:
  - id: mvx-sensitive-no-owner-check
    languages: [rust]
    message: "Sensitive operation without owner check. Add #[only_owner] or explicit verification."
    severity: ERROR
    patterns:
      - pattern: |
          #[endpoint]
          fn $FUNC(&self, $...PARAMS) {
              <... self.$MAPPER().set(...) ...>
          }
      - pattern-not: |
          #[only_owner]
          #[endpoint]
          fn $FUNC(&self, $...PARAMS) { ... }
      - pattern-not: |
          #[endpoint]
          fn $FUNC(&self, $...PARAMS) {
              <... self.blockchain().get_owner_address() ...>
          }
      - metavariable-regex:
          metavariable: $MAPPER
          regex: "(admin|owner|config|fee|rate)"

4.3 Advanced Patterns

Callback Without Error Handling

rules:
  - id: mvx-callback-no-error-handling
    languages: [rust]
    message: "Callback does not handle error case. Async call failures will silently proceed."
    severity: ERROR
    patterns:
      - pattern: |
          #[callback]
          fn $FUNC(&self, $...PARAMS) {
              $...BODY
          }
      - pattern-not: |
          #[callback]
          fn $FUNC(&self, #[call_result] $RESULT: ManagedAsyncCallResult<$TYPE>) {
              ...
          }

Unbounded Iteration

rules:
  - id: mvx-unbounded-iteration
    languages: [rust]
    message: "Iterating over storage mapper without bounds. Can cause DoS via gas exhaustion."
    severity: ERROR
    pattern-either:
      - pattern: self.$MAPPER().iter()
      - pattern: |
          for $ITEM in self.$MAPPER().iter() {
              ...
          }
    metadata:
      category: security
      subcategory: dos
      cwe: "CWE-400: Uncontrolled Resource Consumption"

Storage Key Collision Risk

rules:
  - id: mvx-storage-key-short
    languages: [rust]
    message: "Storage key is very short, increasing collision risk. Use descriptive keys."
    severity: WARNING
    patterns:
      - pattern: '#[storage_mapper("$KEY")]'
      - metavariable-regex:
          metavariable: $KEY
          regex: "^.{1,3}$"

Reentrancy Pattern Detection

rules:
  - id: mvx-reentrancy-risk
    languages: [rust]
    message: "External call before state update. Follow Checks-Effects-Interactions pattern."
    severity: ERROR
    patterns:
      - pattern: |
          fn $FUNC(&self, $...PARAMS) {
              ...
              self.send().$SEND_METHOD(...);
              ...
              self.$STORAGE().set(...);
              ...
          }
      - pattern: |
          fn $FUNC(&self, $...PARAMS) {
              ...
              self.tx().to(...).transfer();
              ...
              self.$STORAGE().set(...);
              ...
          }

4.4 Creating Rules from Findings

Workflow

1. Find a bug manually during audit
2. Abstract the pattern - what makes this a bug?
3. Write a Semgrep rule to catch similar issues
4. Test on the codebase - find all variants
5. Refine to reduce false positives

Example: From Bug to Rule

Bug Found:

#[endpoint]
fn withdraw(&self, amount: BigUint) {
    let caller = self.blockchain().get_caller();
    self.send().direct_egld(&caller, &amount);  // Sends before balance check!
    self.balances(&caller).update(|b| *b -= &amount);  // Can underflow
}

Rule Created:

rules:
  - id: mvx-withdraw-pattern-unsafe
    languages: [rust]
    message: "Withdrawal sends funds before updating balance. Risk of reentrancy and underflow."
    severity: ERROR
    patterns:
      - pattern: |
          fn $FUNC(&self, $...PARAMS) {
              ...
              self.send().$METHOD(...);
              ...
              self.$BALANCE(...).update(|$B| *$B -= ...);
              ...
          }
      - pattern: |
          fn $FUNC(&self, $...PARAMS) {
              ...
              self.tx().to(...).transfer();
              ...
              self.$BALANCE(...).update(|$B| *$B -= ...);
              ...
          }

4.5 Running Semgrep

# Run single rule
semgrep --config rules/mvx-unsafe-arithmetic.yaml src/

# Run all rules in directory
semgrep --config rules/ src/

# Output JSON for processing
semgrep --config rules/ --json -o results.json src/

# Ignore test files
semgrep --config rules/ --exclude="*_test.rs" --exclude="tests/" src/

CI/CD Integration

# GitHub Actions example
- name: Run Semgrep
  uses: returntocorp/semgrep-action@v1
  with:
    config: >-
      rules/mvx-security.yaml
      rules/mvx-best-practices.yaml

4.6 Rule Library Organization

semgrep-rules/
├── security/
│   ├── mvx-arithmetic.yaml      # Overflow/underflow
│   ├── mvx-access-control.yaml  # Auth issues
│   ├── mvx-reentrancy.yaml      # CEI violations
│   └── mvx-input-validation.yaml
├── best-practices/
│   ├── mvx-storage.yaml         # Storage patterns
│   ├── mvx-gas.yaml             # Gas optimization
│   └── mvx-error-handling.yaml
└── style/
    ├── mvx-naming.yaml          # Naming conventions
    └── mvx-documentation.yaml   # Doc requirements

4.7 Testing Rules

Test File Format

# test/mvx-unsafe-unwrap.test.yaml
rules:
  - id: mvx-unsafe-unwrap
    # ... rule definition ...

# Test cases
test_cases:
  - name: "Should match unwrap"
    code: |
      fn test() {
          let x = some_option.unwrap();
      }
    should_match: true

  - name: "Should not match unwrap_or_else"
    code: |
      fn test() {
          let x = some_option.unwrap_or_else(|| sc_panic!("Error"));
      }
    should_match: false

Running Tests

semgrep --test rules/

4.8 Best Practices for Rule Writing

1. Start specific, then generalize: Begin with exact pattern, relax constraints carefully
2. Include fix suggestions: Use fix: field when automated fixes are safe
3. Document the "why": Message should explain impact, not just what's detected
4. Include CWE references: Link to standard vulnerability classifications
5. Test with real codebases: Validate against actual MultiversX projects
6. Version your rules: Rules evolve as framework APIs change
7. Categorize by severity: ERROR for security, WARNING for best practices