PickSkill
← Back

mass-assignment-anti-pattern

Security anti-pattern for mass assignment vulnerabilities (CWE-915). Use when generating or reviewing code that creates or updates objects from user input, form handling, or API request processing. Detects uncontrolled property binding enabling privilege escalation.

SKILL.md
Rendered from GitHub raw
View raw ↗

name: "mass-assignment-anti-pattern" description: "Security anti-pattern for mass assignment vulnerabilities (CWE-915). Use when generating or reviewing code that creates or updates objects from user input, form handling, or API request processing. Detects uncontrolled property binding enabling privilege escalation."

Mass Assignment Anti-Pattern

Severity: High

Summary

Mass assignment (autobinding) occurs when frameworks automatically bind HTTP parameters to object properties without filtering. Attackers inject unauthorized properties (isAdmin: true) to escalate privileges or modify protected fields. This vulnerability enables complete access control bypass through parameter injection.

The Anti-Pattern

Never use user-provided data dictionaries to update models without filtering for allowed properties. Use explicit allowlists.

BAD Code Example

# VULNERABLE: Incoming request data used directly to update user model
from flask import request
from db import User, session
 
@app.route("/api/users/me", methods=["POST"])
def update_profile():
    # User already authenticated
    user = get_current_user()
 
    # Attacker crafts JSON body:
    # {
    #   "email": "new.email@example.com",
    #   "is_admin": true
    # }
    request_data = request.get_json()
 
    # ORMs allow updating objects from dictionaries
    # If User model has `is_admin` property, it updates here
    for key, value in request_data.items():
        setattr(user, key, value) # Direct, unsafe assignment
 
    session.commit()
    return {"message": "Profile updated."}
 
# Attacker just became administrator

GOOD Code Example

# SECURE: Use DTO or explicit allowlist to control updatable fields
from flask import request
from db import User, session
 
# Option 1: Allowlist of fields
ALLOWED_UPDATE_FIELDS = {"email", "first_name", "last_name"}
 
@app.route("/api/users/me", methods=["POST"])
def update_profile_allowlist():
    user = get_current_user()
    request_data = request.get_json()
 
    for key, value in request_data.items():
        # Update only if in explicit allowlist
        if key in ALLOWED_UPDATE_FIELDS:
            setattr(user, key, value)
 
    session.commit()
    return {"message": "Profile updated."}
 
 
# Option 2 (Better): Use DTO or schema for validation
from pydantic import BaseModel, EmailStr
 
class UserUpdateDTO(BaseModel):
    # Defines *only* submittable fields
    # `is_admin` not included, can't be set by user
    email: EmailStr
    first_name: str
    last_name: str
 
@app.route("/api/users/me/dto", methods=["POST"])
def update_profile_dto():
    user = get_current_user()
    try:
        # Pydantic raises validation error if extra fields present
        update_data = UserUpdateDTO(**request.get_json())
    except ValidationError as e:
        return {"error": str(e)}, 400
 
    user.email = update_data.email
    user.first_name = update_data.first_name
    user.last_name = update_data.last_name
    session.commit()
    return {"message": "Profile updated."}

Detection

  • Find direct model updates from request data: Grep for unsafe binding:
    • rg 'setattr.*request\.|\.update\(request\.' --type py
    • rg 'Object\.assign.*req\.body|\.save\(req\.body' --type js
    • rg 'BeanUtils\.copyProperties|ModelMapper' --type java
  • Identify blocklist approaches (insecure): Find key deletion patterns:
    • rg 'del.*\[.*(admin|role|permission)|\.pop\(.*(admin|role)' --type py
    • rg 'delete.*\.(isAdmin|role)|omit\(' --type js
    • Blocklists are insecure - search for allowlist patterns instead
  • Test for mass assignment: Send malicious parameters:
    • curl -X POST /api/users -d '{"email":"test@example.com","isAdmin":true}'
    • Try: isAdmin, role, permissions, accountBalance, verified
  • Check for DTO usage: Verify proper input validation:
    • rg 'class.*DTO|@Valid|validator\.validate' --type py --type java

Prevention

  • Use allowlists: Never use blocklists. Always use allowlists of user-settable properties
  • Use DTOs or schemas: Strictly define expected request body. Most robust solution
  • Set sensitive properties explicitly: Outside mass assignment (e.g., new_user.is_admin = False)
  • Know framework features: Some frameworks include mass assignment protections. Use them correctly

References