Exploiting Excessive Data Exposure in API

Exploiting Excessive Data Exposure in API

When to Use

• Testing APIs where the frontend displays a subset of data but the API response includes additional fields
• Assessing mobile application APIs where responses are designed for multiple client types and may contain excess data
• Identifying PII leakage in API responses that include email addresses, phone numbers, SSNs, or payment data not shown in the UI
• Testing GraphQL APIs where clients can request arbitrary fields including sensitive attributes
• Evaluating APIs after microservice refactoring where internal service-to-service data leaks into public endpoints

Do not use without written authorization. Data exposure testing involves capturing and analyzing potentially sensitive personal data.

Most Often Missed & How to Confirm

• Nested and embedded objects: sensitive fields hide in order.buyer.ssn, comment.author.email, team.members[].last_login_ip - recursively flatten; don't stop at top-level keys.
• Pagination and list-vs-detail drift: list endpoints sometimes return more (or different) fields than the detail view, and page 2+ may differ from page 1.
• Alternate representations: ?fields=, ?expand=, ?include=, .json vs .xml, mobile User-Agent, and API-version headers can unlock extra fields.
• GraphQL field enumeration: request sensitive fields directly even with introspection off (guess passwordHash, mfaSecret, fragments on shared types).
• Error/debug leakage: malformed input often returns stack traces, SQL, internal hostnames, and X-* debug headers.

How to confirm a hit (avoid false negatives): prove the field reaches the client by capturing the raw HTTP response in a proxy, not the rendered UI - the value must be present in the response body/headers and be data the account should not see. Don't conclude negative until you've tried: nested objects, list endpoints, all pagination pages, fields/expand params, GraphQL, error responses, and a second lower-privileged account to confirm it's truly excess exposure.

Prerequisites

• Written authorization specifying target API endpoints and scope
• Burp Suite Professional or mitmproxy configured as intercepting proxy
• Two test accounts at different privilege levels (regular user and admin)
• Browser developer tools or mobile proxy setup for traffic capture
• Python 3.10+ with requests and json libraries
• API documentation (OpenAPI spec) for comparison against actual responses

Legal Notice: This skill is for authorized security testing and educational purposes only. Unauthorized use against systems you do not own or have written permission to test is illegal and may violate computer fraud laws.

Workflow

Step 1: Response Schema Discovery

Compare documented API responses with actual responses:

import requests
import json

BASE_URL = "https://target-api.example.com/api/v1"
headers = {"Authorization": "Bearer <user_token>", "Content-Type": "application/json"}

# Fetch a resource and analyze all returned fields
endpoints_to_test = [
    ("GET", "/users/me", None),
    ("GET", "/users/me/orders", None),
    ("GET", "/products", None),
    ("GET", "/users/me/settings", None),
    ("GET", "/transactions", None),
]

for method, path, body in endpoints_to_test:
    resp = requests.request(method, f"{BASE_URL}{path}", headers=headers, json=body)
    if resp.status_code == 200:
        data = resp.json()
        # Recursively extract all field names
        def extract_fields(obj, prefix=""):
            fields = []
            if isinstance(obj, dict):
                for k, v in obj.items():
                    full_key = f"{prefix}.{k}" if prefix else k
                    fields.append(full_key)
                    fields.extend(extract_fields(v, full_key))
            elif isinstance(obj, list) and obj:
                fields.extend(extract_fields(obj[0], f"{prefix}[]"))
            return fields

        all_fields = extract_fields(data)
        print(f"\n{method} {path} - {len(all_fields)} fields returned:")
        for f in sorted(all_fields):
            print(f"  {f}")

Step 2: Sensitive Data Pattern Detection

Scan API responses for sensitive data patterns:

import re

SENSITIVE_PATTERNS = {
    "email": r'[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}',
    "phone": r'(\+?1?\s?\(?\d{3}\)?[\s.-]?\d{3}[\s.-]?\d{4})',
    "ssn": r'\b\d{3}-\d{2}-\d{4}\b',
    "credit_card": r'\b(?:4[0-9]{12}(?:[0-9]{3})?|5[1-5][0-9]{14}|3[47][0-9]{13})\b',
    "password_hash": r'\$2[aby]?\$\d{2}\$[./A-Za-z0-9]{53}',
    "api_key": r'(?:api[_-]?key|apikey)["\s:=]+["\']?([a-zA-Z0-9_\-]{20,})',
    "internal_ip": r'\b(?:10\.\d{1,3}|172\.(?:1[6-9]|2\d|3[01])|192\.168)\.\d{1,3}\.\d{1,3}\b',
    "aws_key": r'AKIA[0-9A-Z]{16}',
    "jwt_token": r'eyJ[A-Za-z0-9_-]+\.eyJ[A-Za-z0-9_-]+\.[A-Za-z0-9_-]+',
    "uuid": r'[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}',
}

SENSITIVE_FIELD_NAMES = [
    "password", "password_hash", "secret", "token", "ssn", "social_security",
    "credit_card", "card_number", "cvv", "pin", "private_key", "api_key",
    "internal_id", "debug", "trace", "stack_trace", "created_by_ip",
    "last_login_ip", "salt", "session_id", "refresh_token", "mfa_secret",
    "date_of_birth", "bank_account", "routing_number", "tax_id"
]

def scan_response(endpoint, response_text):
    findings = []
    # Check for sensitive data patterns in values
    for pattern_name, pattern in SENSITIVE_PATTERNS.items():
        matches = re.findall(pattern, response_text)
        if matches:
            findings.append({
                "endpoint": endpoint,
                "type": "sensitive_value",
                "pattern": pattern_name,
                "count": len(matches),
                "sample": matches[0][:20] + "..." if len(matches[0]) > 20 else matches[0]
            })

    # Check for sensitive field names
    response_lower = response_text.lower()
    for field in SENSITIVE_FIELD_NAMES:
        if f'"{field}"' in response_lower or f"'{field}'" in response_lower:
            findings.append({
                "endpoint": endpoint,
                "type": "sensitive_field",
                "field_name": field
            })

    return findings

# Scan all endpoint responses
for method, path, body in endpoints_to_test:
    resp = requests.request(method, f"{BASE_URL}{path}", headers=headers, json=body)
    if resp.status_code == 200:
        findings = scan_response(f"{method} {path}", resp.text)
        for f in findings:
            print(f"[FINDING] {f['endpoint']}: {f['type']} - {f.get('pattern', f.get('field_name'))}")

Step 3: Compare UI Display vs API Response

# Fields the UI shows (observed from the frontend application)
ui_displayed_fields = {
    "/users/me": {"name", "email", "avatar_url", "role"},
    "/users/me/orders": {"order_id", "date", "status", "total"},
    "/products": {"id", "name", "price", "image_url", "description"},
}

# Fields the API actually returns
for method, path, body in endpoints_to_test:
    resp = requests.request(method, f"{BASE_URL}{path}", headers=headers, json=body)
    if resp.status_code == 200:
        data = resp.json()
        if isinstance(data, list):
            actual_fields = set(data[0].keys()) if data else set()
        elif isinstance(data, dict):
            # Handle paginated responses
            items_key = next((k for k in data if isinstance(data[k], list)), None)
            if items_key and data[items_key]:
                actual_fields = set(data[items_key][0].keys())
            else:
                actual_fields = set(data.keys())
        else:
            continue

        expected = ui_displayed_fields.get(path, set())
        excess = actual_fields - expected
        if excess:
            print(f"\n{method} {path} - EXCESS FIELDS (not shown in UI):")
            for field in sorted(excess):
                print(f"  - {field}")

Step 4: Test User Object Exposure in Related Endpoints

# Many APIs embed full user objects in responses for orders, comments, etc.
endpoints_with_user_objects = [
    "/orders",          # Each order may include full seller/buyer profile
    "/comments",        # Comments may include full author profile
    "/reviews",         # Reviews may expose reviewer details
    "/transactions",    # Transactions may include counterparty info
    "/team/members",    # Team listing may expose excessive member data
]

for path in endpoints_with_user_objects:
    resp = requests.get(f"{BASE_URL}{path}", headers=headers)
    if resp.status_code == 200:
        text = resp.text
        # Check for user data leakage in nested objects
        user_fields_found = []
        for field in ["password_hash", "last_login_ip", "mfa_enabled", "phone_number",
                      "date_of_birth", "ssn", "internal_notes", "salary", "address"]:
            if f'"{field}"' in text:
                user_fields_found.append(field)
        if user_fields_found:
            print(f"[EXCESSIVE] {path} exposes user fields: {user_fields_found}")

Step 5: GraphQL Over-Fetching Analysis

# GraphQL allows clients to request any available field
GRAPHQL_URL = f"{BASE_URL}/graphql"

# Introspection query to discover all fields on User type
introspection = {
    "query": """
    {
      __type(name: "User") {
        fields {
          name
          type {
            name
            kind
          }
        }
      }
    }
    """
}

resp = requests.post(GRAPHQL_URL, headers=headers, json=introspection)
if resp.status_code == 200:
    fields = resp.json().get("data", {}).get("__type", {}).get("fields", [])
    print("Available User fields via GraphQL:")
    for f in fields:
        sensitivity = "SENSITIVE" if f["name"] in SENSITIVE_FIELD_NAMES else "normal"
        print(f"  {f['name']} ({f['type']['name']}) [{sensitivity}]")

# Try to query sensitive fields
sensitive_query = {
    "query": """
    query {
      users {
        id
        email
        passwordHash
        socialSecurityNumber
        internalNotes
        lastLoginIp
        mfaSecret
        apiKey
      }
    }
    """
}
resp = requests.post(GRAPHQL_URL, headers=headers, json=sensitive_query)
if resp.status_code == 200 and "errors" not in resp.json():
    print("[CRITICAL] GraphQL exposes sensitive user fields without restriction")

Step 6: Debug and Internal Data Leakage

# Test for debug information in responses
debug_headers_to_check = [
    "X-Debug-Token", "X-Debug-Info", "Server", "X-Powered-By",
    "X-Request-Id", "X-Correlation-Id", "X-Backend-Server",
    "X-Runtime", "X-Version", "X-Build-Version"
]

resp = requests.get(f"{BASE_URL}/users/me", headers=headers)
for h in debug_headers_to_check:
    if h.lower() in {k.lower(): v for k, v in resp.headers.items()}:
        print(f"[INFO LEAK] Header {h}: {resp.headers.get(h)}")

# Test error responses for stack traces
error_payloads = [
    ("GET", "/users/invalid-id-format", None),
    ("POST", "/orders", {"invalid": "payload"}),
    ("GET", "/users/-1", None),
    ("GET", "/users/0", None),
]

for method, path, body in error_payloads:
    resp = requests.request(method, f"{BASE_URL}{path}", headers=headers, json=body)
    if resp.status_code >= 400:
        text = resp.text.lower()
        if any(kw in text for kw in ["stack trace", "traceback", "at com.", "at org.",
                                      "file \"", "line ", "exception", "sql", "query"]):
            print(f"[DEBUG LEAK] {method} {path} -> {resp.status_code}: Contains stack trace or query info")

Key Concepts

Tools & Systems

• Burp Suite Professional: Intercept API responses and use the Comparer tool to diff expected vs actual response schemas
• mitmproxy: Scriptable proxy for automated response analysis with Python-based content inspection scripts
• OWASP ZAP: Passive scanner detects information disclosure in headers, error messages, and response bodies
• Postman: Compare documented response schemas against actual API responses using test scripts
• jq: Command-line JSON processor for extracting and analyzing specific fields from API responses

Common Scenarios

Scenario: Mobile Banking API Data Exposure Assessment

Context: A mobile banking application's API returns full account objects to the mobile client, which only displays account nickname and balance. The API is accessed by both iOS and Android apps and a web portal.

Approach:

1. Configure mitmproxy on a test device and authenticate as the test user
2. Capture all API responses during a complete user session (login, view accounts, transfer, logout)
3. Analyze GET /api/v1/accounts response: UI shows 4 fields but API returns 23 fields
4. Discover that the API returns routing_number, account_holder_ssn_last4, internal_risk_score, kyc_verification_status, and linked_external_accounts - none shown in UI
5. Analyze GET /api/v1/transactions response: API returns merchant_id, terminal_id, authorization_code, processor_response fields not needed by the client
6. Check GET /api/v1/users/me: API returns last_login_ip, mfa_backup_codes_remaining, account_officer_name, and credit_score_band
7. Test error responses: POST /api/v1/transfers with invalid payload returns SQL table name in error message

Pitfalls:

• Only checking top-level fields and missing sensitive data in deeply nested objects
• Not testing paginated responses where subsequent pages may include different fields
• Ignoring response headers that may leak server version, backend technology, or internal routing information
• Missing data exposure in error responses which often contain stack traces, SQL queries, or internal paths
• Assuming that HTTPS encryption prevents data exposure (it protects in transit, not from the authenticated client)

Output Format

## Finding: Excessive Data Exposure in Account and Transaction APIs

**ID**: API-DATA-001
**Severity**: High (CVSS 7.1)
**OWASP API**: API3:2023 - Broken Object Property Level Authorization
**Affected Endpoints**:
  - GET /api/v1/accounts
  - GET /api/v1/transactions
  - GET /api/v1/users/me

**Description**:
The API returns full database objects to the client, including sensitive fields
that are not displayed in the mobile application UI. The mobile app filters
these fields client-side, but they are fully accessible by intercepting the
API response. This exposes SSN fragments, internal risk scores, and KYC
verification data for any authenticated user.

**Excess Fields Discovered**:
- /accounts: routing_number, account_holder_ssn_last4, internal_risk_score,
  kyc_verification_status, linked_external_accounts (18 excess fields total)
- /transactions: merchant_id, terminal_id, authorization_code,
  processor_response (12 excess fields total)
- /users/me: last_login_ip, mfa_backup_codes_remaining, credit_score_band

**Impact**:
An authenticated user can extract sensitive financial data, internal risk
assessments, and PII for their own account that the application is not
intended to reveal. Combined with BOLA vulnerabilities, this data could
be extracted for all users.

**Remediation**:
1. Implement server-side response filtering using DTOs/view models that only include fields needed by the client
2. Use GraphQL field-level authorization or REST response schemas per endpoint per role
3. Remove sensitive fields from API responses at the serialization layer
4. Implement response schema validation in the API gateway to strip undocumented fields
5. Add automated tests that verify response schemas match documentation