Home > CAPEC List > CAPEC-34: HTTP Response Splitting (Version 2.10)  

CAPEC-34: HTTP Response Splitting

 
HTTP Response Splitting
Definition in a New Window Definition in a New Window
Attack Pattern ID: 34
Abstraction: Detailed
Status: Draft
Completeness: Complete
Presentation Filter:
+ Summary

This attack uses a maliciously-crafted HTTP request in order to cause a vulnerable web server to respond with an HTTP response stream that will be interpreted by the client as two separate responses instead of one. This is possible when user-controlled input is used unvalidated as part of the response headers. The target software, the client, will interpret the injected header as being a response to a second request, thereby causing the maliciously-crafted contents be displayed and possibly cached.

To achieve HTTP Response Splitting on a vulnerable web server, the attacker:

+ Attack Execution Flow
Explore
  1. Spider:

    Using a browser or an automated tool, an attacker follows all public links on a web site. He records all the links, the forms and all potential user-controllable input points for the web application.

    Attack Step Techniques

    IDAttack Step Technique DescriptionEnvironments
    1

    Use a spidering tool to follow and record all links and analyze the web pages to find entry points. Make special note of any links that include parameters in the URL, forms found in the pages (like file upload, etc.).

    env-Web
    2

    Use a proxy tool to record all links visited during a manual traversal of the web application.

    env-Web
    3

    Use a browser to manually explore the website and analyze how it is constructed. Many browsers' plugins are available to facilitate the analysis or automate the discovery.

    env-Web

    Indicators

    IDTypeIndicator DescriptionEnvironments
    1Positive

    Inputs are transported through HTTP

    env-Web
    2Positive

    The application uses redirection techniques (HTTP Location, etc.)

    env-Web
    3Inconclusive

    Using URL rewriting, parameters may be part of the URL path.

    env-Web
    4Inconclusive

    No parameters appear to be used on the current page. Even though none appear, the web application may still use them if they are provided.

    env-Web
    5Negative

    Applications that have only static pages or that simply present information without accepting input are unlikely to be susceptible.

    env-Web

    Outcomes

    IDTypeOutcome Description
    1Success
    A list of user-controllable input entry points is created by the attacker.

    Security Controls

    IDTypeSecurity Control Description
    1Detective
    Monitor velocity of page fetching in web logs. Humans who view a page and select a link from it will click far slower and far less regularly than tools. Tools make requests very quickly and the requests are typically spaced apart regularly (e.g. 0.8 seconds between them).
    2Detective
    Create links on some pages that are visually hidden from web browsers. Using iframes, images, or other HTML techniques, the links can be hidden from web browsing humans, but visible to spiders and programs. A request for the page, then, becomes a good predictor of an automated tool probing the application.
    3Preventative
    Use CAPTCHA to prevent the use of the application by an automated tool.
    4Preventative
    Actively monitor the application and either deny or redirect requests from origins that appear to be automated.
Experiment
  1. Attempt variations on input parameters:

    The attacker injects the entry points identified in the Explore Phase with response splitting syntax and variations of payloads to be acted on in the additional response. He records all the responses from the server that include unmodified versions of his payload.

    Attack Step Techniques

    IDAttack Step Technique DescriptionEnvironments
    1

    Use CR\LF characters (encoded or not) in the payloads in order to see if the HTTP header can be split.

    env-Web
    2

    Use a proxy tool to record the HTTP responses headers.

    env-Web

    Indicators

    IDTypeIndicator DescriptionEnvironments
    1Positive

    The web server uses unvalidated user-controlled input as part of the response headers

    env-Web

    Outcomes

    IDTypeOutcome Description
    1Success
    The CR\LF characters are passed in the HTTP header and two responses are generated for a single request.
    2Failure
    All CR\LF characters are consistently re-encoded or stripped before being written in the HTTP header
    3Inconclusive
    The size of the payload is being limited by the server-side application.
    4Inconclusive
    Some sensitive characters are consistently encoded, but others are not.

    Security Controls

    IDTypeSecurity Control Description
    1Detective
    Monitor input to web servers (not only GET, but all in the inputs), application servers, and other HTTP infrastructure (e.g., load balancers). Alert on CR\LF characters.
    2Preventative
    Do not use user-controllable inputs in HTTP headers
    3Preventative
    Filter CR/LF syntax out of any user-controllable input utilized in HTTP headers.
    4Preventative
    Actively monitor the application and either deny or redirect requests from origins that appear to be generating HTTP Response Splitting attacks.
Exploit
  1. Cross-Site Scripting:

    As the attacker succeeds in exploiting the vulnerability, he can choose to attack the user with Cross-Site Scripting. The possible outcomes of such an attack are described in the Cross-Site Scripting related attack patterns.

    Attack Step Techniques

    IDAttack Step Technique DescriptionEnvironments
    1

    Inject cross-site scripting payload preceded by response splitting syntax (CR/LF) into user-controllable input identified as vulnerable in the Experiment Phase.

    env-Web

    Outcomes

    IDTypeOutcome Description
    1Success
    The malicious script is executed within the user's context.

    Security Controls

    IDTypeSecurity Control Description
    1Detective
    Monitor server logs for consecutive suspicious HTTP request
    2Preventative
    Apply appropriate input validation to filter all user-controllable input of scripting syntax
    3Preventative
    Appropriately encode all browser output to avoid scripting syntax
  2. Cache poisoning:

    Attack Step Techniques

    IDAttack Step Technique DescriptionEnvironments
    1

    The attacker decides to target the cache server by forging new responses. The server will then cache the second request and response. The cached response has most likely an attack vector like Cross-Site Scripting; this attack will then be serve to many clients due to the caching system.

    env-Web

    Indicators

    IDTypeIndicator DescriptionEnvironments
    1Positive

    System performs caching of HTTP responses

    env-Web

    Outcomes

    IDTypeOutcome Description
    1Success
    The attacker gets the users to be served with this cached malicious HTTP response.

    Security Controls

    IDTypeSecurity Control Description
    1Detective
    Monitor server logs for consecutive suspicious HTTP requests.
    2Preventative
    Apply appropriate input validation to filter all user-controllable input of scripting syntax
    3Preventative
    Appropriately encode all browser output to avoid scripting syntax
+ Attack Prerequisites
  • User-controlled input used as part of HTTP header

  • Ability of attacker to inject custom strings in HTTP header

  • Insufficient input validation in application to check for input sanity before using it as part of response header

+ Typical Severity

High

+ Typical Likelihood of Exploit

Likelihood: Medium

+ Methods of Attack
  • Injection
  • Protocol Manipulation
+ Examples-Instances

Description

In the PHP 5 session extension mechanism, a user-supplied session ID is sent back to the user within the Set-Cookie HTTP header. Since the contents of the user-supplied session ID are not validated, it is possible to inject arbitrary HTTP headers into the response body. This immediately enables HTTP Response Splitting by simply terminating the HTTP response header from within the session ID used in the Set-Cookie directive.

Related Vulnerabilities

+ Attacker Skills or Knowledge Required

Skill or Knowledge Level: High

The attacker needs to have a solid understanding of the HTTP protocol and HTTP headers and must be able to craft and inject requests to elicit the split responses.

+ Resources Required

None

+ Probing Techniques

With available source code, the attacker can see whether user input is validated or not before being used as part of output. This can also be achieved with static code analysis tools

If source code is not available, the attacker can try injecting a CR-LF sequence (usually encoded as %0d%0a in the input) and use a proxy such as Paros to observe the response. If the resulting injection causes an invalid request, the web server may also indicate the protocol error.

+ Indicators-Warnings of Attack

The only indicators are multiple responses to a single request in the web logs. However, this is difficult to notice in the absence of an application filter proxy or a log analyzer. There are no indicators for the client

+ Solutions and Mitigations

To avoid HTTP Response Splitting, the application must not rely on user-controllable input to form part of its output response stream. Specifically, response splitting occurs due to injection of CR-LF sequences and additional headers. All data arriving from the user and being used as part of HTTP response headers must be subjected to strict validation that performs simple character-based as well as semantic filtering to strip it of malicious character sequences and headers.

+ Attack Motivation-Consequences
ScopeTechnical ImpactNote
Confidentiality
Integrity
Availability
Execute unauthorized code or commands
Run Arbitrary Code
Confidentiality
Access_Control
Authorization
Gain privileges / assume identity
+ Injection Vector

User-controllable input that forms part of output HTTP response headers

+ Payload

Encoded HTTP header and data separated by appropriate CR-LF sequences. The injected data must consist of legitimate and well-formed HTTP headers as well as required script to be included as HTML body.

+ Activation Zone

API calls in the application that set output response headers.

+ Payload Activation Impact

The impact of payload activation is that two distinct HTTP responses are issued to the target, which interprets the first as response to a supposedly valid request and the second, which causes the actual attack, to be a response to a second dummy request issued by the attacker.

+ Relevant Security Requirements

All client-supplied input must be validated through filtering and all output must be properly escaped.

+ Purposes
  • Penetration
+ CIA Impact
Confidentiality Impact: HighIntegrity Impact: HighAvailability Impact: Low
+ Technical Context
Architectural Paradigms
Client-Server
n-Tier
Frameworks
All
Platforms
All
Languages
All
+ References
[R.34.1] [REF-2] G. Hoglund and G. McGraw. "Exploiting Software: How to Break Code". Addison-Wesley. February 2004.
[R.34.2] [REF-3] "Common Weakness Enumeration (CWE)". CWE-113 - HTTP Response Splitting. Draft. The MITRE Corporation. 2007. <http://cwe.mitre.org/data/definitions/113.html>.
[R.34.3] [REF-3] "Common Weakness Enumeration (CWE)". CWE-74 - Injection. Draft. The MITRE Corporation. 2007. <http://cwe.mitre.org/data/definitions/74.html>.
+ Content History
Submissions
SubmitterOrganizationDateSource
CAPEC Content TeamThe MITRE Corporation2014-06-23Internal_CAPEC_Team
More information is available — Please select a different filter.
Page Last Updated or Reviewed: May 01, 2017