Race Condition (TOCTOU) Exploitation

When to Use

• When testing critical business logic that updates a finite state (balances, stock/inventory, one-time coupons, likes/votes).
• When a single action should be permitted once per user, but you suspect the back-end database locking mechanism (e.g., ACID transactions or Mutexes) is flawed.
• When an application executes a "Check" (Do they have enough money?) followed by a delayed "Use" (Deduct the money). Note: This split-second window between check and use is the target.

Prerequisites

• Authorized scope and target URLs from bug bounty program
• Burp Suite Professional (or Community) configured with browser proxy
• Familiarity with OWASP Top 10 and common web vulnerability classes
• SecLists wordlists for fuzzing and enumeration

Workflow

Phase 1: Identifying Race Windows

# Concept: You need to identify endpoints that perform sequence-dependent logic.
# A classic example: Applying a $50 discount code to a cart.

# The Flawed Backend Logic (Conceptual):
# 1. Check if 'DISCOUNT_CODE' is unused for User A. (Time of Check)
# 2. Add $50 discount to User A's cart.
# 3. Mark 'DISCOUNT_CODE' as used for User A. (Time of Use)

# Goal: Send 20 concurrent requests. If requests 2-20 hit the server BEFORE 
# request 1 marks the code as used in step 3, the server will apply the discount 20 times.

Phase 2: Single-Packet Attack (Burp Suite Pro)

# Concept: Network latency and jitter often ruin race conditions because your 
# 20 requests arrive at slightly different milliseconds.
# Modern technique: HTTP/2 Single-Packet Attack (sending multiple requests within a single TCP packet).

# 1. Send the target request to Burp Repeater (e.g., POST /api/apply_coupon).
# 2. Group the requests: 
#    - Duplicate the tab 20 times.
#    - Select all 20 tabs, right-click -> "Add tab to group" -> Create new group.
# 3. Execution:
#    - Click the dropdown arrow next to the "Send" button.
#    - Select "Send group in parallel (last-byte sync)".
#    - Click Send.

# 4. Analysis:
# If you receive multiple "200 OK - Coupon Applied" responses, you successfully raced the endpoint.

Phase 3: Turbo Intruder (High-Velocity Exploitation)

# Concept: If Single-Packet Attack isn't suitable, use PortSwigger's Turbo Intruder extension
# to blast hundreds of requests utilizing a custom Python pipeline for raw socket speed.

# 1. Send request to Turbo Intruder.
# 2. Use the standard race condition script:

def queueRequests(target, wordlists):
    engine = RequestEngine(endpoint=target.endpoint,
                           concurrentConnections=30,
                           requestsPerConnection=100,
                           pipeline=False)

    request = target.req
    
    # Queue up 30 requests to be released simultaneously
    for i in range(30):
        engine.queue(request, gate='race1')

    # Release all 30 requests perfectly synchronized
    engine.openGate('race1')

def handleResponse(req, interesting):
    table.add(req)

# 3. Click "Attack".

Phase 4: Multi-Endpoint Race Conditions

# Concept: Racing isn't just for duplicating single actions. It applies to conflicting actions.
# Example: Bypassing an email validation restriction.

# Endpoint A: POST /update_email (requires validation)
# Endpoint B: POST /checkout (requires verified email)

# The Flaw:
# 1. Attacker sets email to verified@trusted.com.
# 2. Attacker initiates checkout.
# 3. WHILE checkout is processing, attacker races a request to change email back to attacker@evil.com.
# 4. Checkout completes using attacker@evil.com, bypassing validation.

# Tooling: Group differing tabs in Burp Repeater and "Send in parallel".

Decision Point 🔀

flowchart TD
    A[Identify State-Changing Endpoint] --> B{Does application rely on quotas/limits?}
    B -->|Yes| C[Queue 20 identical requests]
    C --> D[Send via HTTP/2 Single Packet Sync]
    D --> E{Did the action succeed multiple times?}
    E -->|Yes| F[Exploit TOCTOU! Document impact.]
    E -->|No| G[Try Turbo Intruder or test different endpoints]
    B -->|No| H[Test for multi-endpoint logical races]

🔵 Blue Team Detection & Defense

• Pessimistic Locking: In a relational database, enforce SELECT ... FOR UPDATE when reading the user's balance or coupon state. This locks the database row until the entire transaction (Check AND Use) commit is fully resolved, forcing concurrent requests to queue sequentially.
• Atomic Operations: Use native database atomic operations rather than fetching, modifying, and writing in application code.
  • VULNERABLE: balance = db.get("balance"); db.update(balance - 50)
  • SECURE: UPDATE users SET balance = balance - 50 WHERE balance >= 50
• Rate Limiting (Partial fix): Extremely aggressive rate-limiting can mitigate generic flooding, but will not stop an HTTP/2 Single-Packet attack where 30 requests arrive simultaneously in one network frame.

Key Concepts

Concept	Description
Race Condition	A flaw occurring when the timing of events affects the correctness of a system, specifically when concurrent threads access shared data blindly
TOCTOU	Time-Of-Check to Time-Of-Use; the specific time window between validating a condition and acting upon that condition
Single-Packet Attack	A technique utilizing HTTP/2 multiplexing to place the trailing byte of multiple requests into a single synchronized TCP packet

Output Format

Bug Bounty Report: Financial Exploitation via Race Condition
============================================================
Vulnerability: State-Spanning TOCTOU Race Condition
Severity: Critical (CVSS 9.0)
Target: POST /api/wallet/transfer

Description:
The money transfer API endpoint is vulnerable to a Time-of-Check to Time-of-Use (TOCTOU) race condition. The backend logic checks the user's account balance, but fails to implement pessimistic row locking during the database transaction that subsequently deducts the funds. 

Reproduction Steps:
1. Create an account with a balance of $100.
2. Intercept a transfer request sending $100 to an attacker-controlled account.
3. Send the request to Burp Repeater and duplicate the tab 20 times.
4. Add all 20 tabs to a new Group.
5. In the Send dropdown, select "Send group in parallel (last-byte sync)".
6. Click Send.

Impact:
The server processes multiple requests simultaneously, validating the $100 balance for the first ~5 threads before the deduction occurs. The attacker successfully transferred $500 out of an account containing only $100, driving the database ledger into the negative. This allows infinite infinite monetary theft.

📚 Shared Resources

For cross-cutting methodology applicable to all vulnerability classes, see:

• _shared/references/elite-chaining-strategy.md — Exploit chaining methodology and high-payout chain patterns
• _shared/references/elite-report-writing.md — HackerOne-optimized report writing, CWE quick reference
• _shared/references/real-world-bounties.md — Verified disclosed bounties by vulnerability class

References

• PortSwigger: Race conditions
• DEF CON 31: Smashing the State Machine (Single-Packet Attack)
• OWASP: Testing for Race Conditions