'Waldo' Finds Ways To Abuse HTML5 WebSockets

With new features come new risks, and HTML5 is no exception: Its WebSocket feature opens the door for a wide range of abuses by attackers, researchers will demonstrate next week at Black Hat USA.

Sergey Shekyan and Vaagn Toukharian also will release a hacking tool called Waldo for researchers to test for WebSocket vulnerabilities and potential attack vectors. WebSocket is basically a faster, more efficient way for browsers and Web servers to communicate (bidirectionally); it supports real-time applications, such as customer-support chat features and online gaming, for example. Most major browsers support it.

The problem is that today's firewalls, next-generation firewalls, unified threat management (UTM), and IDS/IPS products are not WebSocket-aware, the researchers say. In other words, they can't detect WebSocket traffic: "Nobody monitors it: We think that's the scariest part," says Shekyan, who is a senior software engineer.

WebSocket could be used for anonymized, TOR-type peer-to-peer communication, for example, as well as a platform for attacks, such as cross-site scripting (XSS) and denial-of-service, the researchers say. The good news is that WebSocket so far is deployed in a small percentage of websites today, only around 1 percent, according to their research.

Waldo is a simple tool based on the Websocketpp server that shows how easy it is to abuse WebSocket on both the client and server side as a malicious communication channel, they say. It currently runs on OS X and most flavors of Linux.

"Our tool is doing malicious things, like delivering commands to a compromised browser to steal cookies and keystrokes, copy documents, and then send them back to us," Shekyan says.

[ HTML5 could help spur SQL injection attacks on client machines, experts say. See New HTML Version Comes With Security Risks Of Its Own. ]

An attacker could launch a XSS exploit against a website to hijack its customer-support chat application, for instance. Many of these chat apps are supplied by a third party. "It's using SSL, and everything seems fine. But the main page is not from the same Web server as the one serving up the chat, and it's using plain HTTP," Shekyan says. An attacker then could execute a man-in-the-middle attack and modify the main page, intercept traffic from the browser, and pose as the customer service rep online, requesting the customer's personal information.

Toukharian, who is a developer for Qualys' Web application scanner, says it's not the WebSocket protocol that's vulnerable: It's the way organizations are implementing it. "All XSS and other vulnerabilities and man-in-the-middle WebSocket connections are weaker from the client side. The problem here is the security design of the entire Web app," he says.

From the client side, an attacker could DoS a server from a single browser because WebSocket supports 1,000 concurrent sessions. "A number of simultaneous browser connections can be opened, which can be a problem," Toukharian says.

The researchers say perimeter security vendors don't consider WebSocket a major threat for now.

They plan to share with Black Hat attendees new data they are gathering on WebSocket use around the Net. "We're going to provide statistics on who's using it ... if it's going over SSL or plain text" and other usage trends, Shekyan says.

While adoption is fairly low today, WebSocket is attractive due to its persistent connection, low overhead, and faster performance, so the concern is that this could become a hot attack vector once adoption rises. And if such attacks are under way today, they wouldn't be detectable from today's security tools, the researchers say.

Have a comment on this story? Please click "Add Your Comment" below. If you'd like to contact Dark Reading's editors directly, send us a message.