Depending on their OAuth implementation, social media sites, such as Google, LinkedIn and Yahoo, could share users' personal information with malicious third-party websites, experts say.
OAuth is an open framework used by social networks to authorize third-party sites seeking profile information from network users who provided permission for the data sharing. Security problems arise when the social network implements OAuth in a way that it cannot determine for certain the third-party site is who it says it is.
Jing Wang, a PhD student in mathematics from Nanyang Technological University in China, described the problem, called a "covert redirect, in a research paper.
Which, if any, major Internet sites are affected is not clear, since it would depend on their OAuth implementation. Wang claimed Facebook is vulnerable, but the site declined to fix the problem immediately.
"Short of forcing every single application on the platform to use a whitelist, (the problem) isn't something that can be accomplished in the short term," Facebook said, according to Wang.
Facebook acknowledged the problem, and said it has instructions on its developers site on how to prevent it.
"We publish guidance to help developers avoid this type of issue that can arise if apps contain certain types of vulnerabilities," a spokesman said.
"Creating a URL whitelist is an effective defense that we include in our developer documentation."
OAuth 2.0, the latest version of the framework, does not require a website to restrict information sharing to only those sites with a uniform resource identifier (URI) on the OAuth whitelist. Sites that choose to skip this requirement, which is allowed in the framework, accept the risk of providing user information to bogus sites.
Despite the option of using the URI whitelist, some experts believe it is still a flaw that the framework does not require it.
"The framework should be written such that there is no option to not specify the redirect URI," Kevin O'Brien, director of product marketing at cloud security vendor CloudLock, said.
To take advantage of OAuth, an attacker would first have to lure a person to a location disguised as a trusted site, experts say. This is typically done through a phishing email with a link to the malicious site.
In logging into the bogus site, visitors would get a drop-down box asking if they want to log in using, for example, their Facebook credentials.
Because of the way OAuth works, the phisher would not get visitors user names and passwords, but they would get whatever personal information they permit Facebook to share, such as name, birth date, email addresses and contact lists.
Attackers could potentially get more information, if they then redirected visitors to another location disguised as a trusted site asking for more information. In redirecting victims, attackers could hide the URL of the bogus site and show the URL of the legitimate site instead.
The threat to companies is not new in that it starts as a typical phishing attack, Brandon Edwards, vice president of labs, at cloud security vendor SilverSky, said.
"The threat seems to be limited to social networking and the impact (to the permissions) that would normally be granted by these apps," Edwards said.
More on single sign-on and identity management:
- What is identity management? Its definition, uses and solutions
- The best identity management advice right now
- What is SAML, what is it used for and how does it work?
- What is OAuth? How the open authorization framework works