OWASP Periodic Table of Vulnerabilities - Brute Force Login

Return to Periodic Table Working View

Root Cause Summary
The application does not prevent attackers from trying many username/password combinations in rapid succession in order to guess account credentials.

Browser / Standards Solution
None

Perimeter Solution
The perimeter should detect brute force login attempts and help prevent denial of service against the login functionality of the application. The perimeter should signal to the application when a username or password is under attack, and accept signals from the application when a certain username or password has been temporarily locked out.

Generic Framework Solution
The framework should track failed login attempts even for non-existent accounts, so that the lockout feature doesn't expose an account enumeration vulnerability.

The framework should track all unique accounts tried unsuccessfully for a particular password, using a hash of the password with a static salt as a key to prevent password exposure. When the threshold is reached, the application should prevent all logins using that password to any account, much like the account lockout.

The framework should be capable of sending notifications to perimeter technologies to help enforce lockouts or global CAPTCHA.

The lockout period should only be temporary, so as to prevent long-term denial of service against a user account. The framework should provide an interface for an administrator or customer service representative to white-list IPs for an account to log in from, in case of a targeted, sustained, long-term denial of service attack against a specific user.

The framework should expose and implement configurable rules about failed login attempts with secure defaults, including:


 * Account lockout
 * Maximum consecutive failures before account lockout (e.g. 5)
 * Initial lockout period (e.g. 5 minutes) per password and per user account
 * Account lockout escalation pattern (e.g. linear, geometric, exponential) - for instance, the lockout period might double after each set of consecutive failures and a different pattern might apply for usernames or passwords
 * Maximum lockout interval (e.g. 24 hours) to put a cap on the escalation pattern
 * Total number of non-existent accounts to track (allows operations to scale the defense based on available infrastructure)
 * URL/file for account lockout error message content
 * Fixed-password, variable-account attack protection
 * Total number of passwords to track (allows operations to scale defense based on available infrastructure)
 * Total number of unique accounts unsuccessfully attempted for a password before locking out the password
 * URL/file for password lockout error message content
 * CAPTCHA
 * Maximum consecutive failures before requiring CAPTCHA (e.g. 3)
 * Number of global authentication attempts per time period before global CAPTCHA is triggered (e.g. 1000 per minute)
 * Global CAPTCHA enforcement period after an event (e.g. 4 hours)
 * CAPTCHA(s) to be used (may select multiple for possible added difficulty) - should support popular CAPTCHAs in use as well as provide an API for custom CAPTCHAs (though not recommended)
 * Authentication delay
 * Maximum consecutive failures before enforcing an authentication delay (e.g. 3)
 * Delay in milliseconds between login attempts after delay is triggered (e.g. 2000)

Custom Framework Solution
None

Custom Code Solution
None

Discussion / Controversy
Perimeter solutions can certainly independently detect brute-force login attacks and even help enforce account lockout, but it is better to maintain control over the lockout rules and account management within the application.

The application may signal to the perimeter when accounts are locked/unlocked or a global CAPTCHA is triggered so it can help prevent a denial-of-service condition against the login functionality.

Framework and perimeter solution designers may wish to collaborate to define an open standard for communicating information about account lockouts, brute force attacks, and anti-automation measures.