Those services allow someone to login to a Web service using, for example, their LinkedIn credentials. It's a convenient way for users to create new accounts on websites by using existing information.
But in one instance, the researchers found they could gain control of accounts at Slashdot.org, Nasdaq.com, Crowdfunder.com and others by abusing LinkedIn's social login mechanism.
Other identity services were also found to be vulnerable to the "SpoofedMe" attack, wrote Or Peles and Roee Hay of IBM Security Systems.
LinkedIn, Amazon and Vasco, all identity providers, have all either fixed or taken measures to prevent such account takeovers, after notification from IBM, the researchers said. But the problem is one that both identity providers and third-party websites using those services should be aware of.
The attack that abuses LinkedIn is demonstrated in a video included in a blog post. The attacker creates an account with LinkedIn, using the victim's email address.
LinkedIn will send a verification email to the victim to ensure the person has control over the address. But for the attacker's purposes, that doesn't matter.
Once the LinkedIn account is created, the attacker goes to Slashdot and uses the social login feature, selecting LinkedIn as the identity provider. Identity providers don't pass along a person's credentials to the third party site, but do transfer information such as an email address.
Slashdot.org matches the email address of the victim that was passed to it by LinkedIn to the existing account, allowing the attacker to control the account. The account could then be used to post malicious links, with people believing a trusted contact posted the content.
The researchers cautioned that the attack only works if the victim doesn't already have an account with an identity provider.
There are two flaws in this approach: Slashdot.org shouldn't trust the email address unless the identity provider knows it has been verified, and the identity provider should wait until the email address has been verified.
LinkedIn was vulnerable when it was using the deprecated 1.0a version of the OAuth protocol for social logins. LinkedIn also uses OAuth 2.0, which stops the authentication process if the email address is unverified, they wrote.
But "it's worth mentioning at the time of this writing, the absolute majority of websites we encountered using LinkedIn as a provider were using the 'deprecated' (and vulnerable) version," the researchers wrote.
The researchers found a similar issue with Amazon's "Sign in with Amazon."
"We found that in addition to allowing us to register an account with an email address we don't own and pass it as part of the social login authentication process, it also allowed us to change the Amazon account email address to another unverified address, making it the new account's primary email address," they wrote.
Amazon's security team has since written more documentation on how local accounts should be linked, including verifying emails, they wrote.
Send news tips and comments to email@example.com. Follow me on Twitter: @jeremy_kirk