Industry Mobilizes on Latest TCP Flaw
The latest security vulnerability discovered in the Transport Control Protocol (TCP) has companies scrambling this week, with Cisco Systems Inc. (Nasdaq: CSCO), Microsoft Corp. (Nasdaq: MSFT), and Sun Microsystems Inc. (Nasdaq: SUNW) all issuing security alerts.
The vulnerability could allow someone to use the Internet Control Message Protocol (ICMP) to launch a denial-of-service (DOS) attack on a router or server. Because the problem is inherent to TCP, it's not isolated to certain companies.
Then again, not all products are affected. Cisco's security advisory includes a table showing which of its products are prone to which types of ICMP attack.
Cisco's advisory notes that no known malicious exploits of the vulnerability have yet cropped up.
The vulnerability isn't new. It was documented in a December Internet Engineering Task Force (IETF) draft by Fernando Gont of the Universidad Tecnológica Nacional / Facultad Regional Haedo in Argentina. But the problem got a spike of public attention with an April 12 alert issued by the National Infrastructure Security Coordination Centre (NISCC), a U.K. government organization that assesses and publicizes threats to network infrastructure. That alert led to Cisco, Microsoft, and Sun issuing security advisories.
Gont's paper is available here.
This is hardly the first problem discovered in TCP, the protocol that rides atop the Internet Protocol (IP) to control Internet traffic. In fact, a TCP vulnerability was one key to Kevin Mitnick's 1994 break-in of programmer Tsutomu Shimomura's network, a case detailed in the books Takedown and The Cyberthief and the Samurai.
ICMP delivers error messages from router to router; it's the messenger telling network elements that a failure has occurred somewhere, for instance. In that case, known as a "hard error," a TCP device will try to terminate the relevent connection (TCP being a connection-oriented protocol). These kinds of reactions can be exploited repeatedly to keep a router in a perpetual state of reset, creating a DOS attack.
The alerts list two other ways to exploit ICMP. One involves devices using Path MTU Discovery, an IETF-defined feature for determining the maximum transmission unit (MTU) of a particular Internet path. These devices can get slowed down through the use of forged ICMP packets related to Path MTU. Finally, it's possible to cause mischief by misusing ICMP Source Quench packets, which are intended to be used for congestion control.
The NISCC believes routers running the Border Gateway Protocol (BGP) are particularly vulnerable to these attacks, because BGP maintains continual TCP connections between routers.
Workarounds can prevent some of the attacks in some types of equipment. The Path MTU exploit can be thwarted if it's possible to turn off Path MTU Discovery, for example. In other cases, there won't be a fix. Cisco notes that its IP phones can't entirely prevent the "hard error" and Source Quench cases, but notes that attacks "can be mitigated by segmenting voice and data through the use of VLAN technologies" and by exercising recommended security precautions.
Gont's paper describes a few ways to partially verify an ICMP message -- an important point, because there is no security check on incoming ICMP packets. One would be to check the TCP sequence number inside the message; if the sequence number doesn't correspond to a packet the router knows is en route, the message can't be valid.
It's been a busy security week for Cisco, as the ICMP problem comes on the heels of Cisco's most recent -- and wholly unrelated -- security advisories (see Cisco Discloses Latest IOS Flaws).
— Craig Matsumoto, Senior Editor, Light Reading
CALLING ALL SECURITY APPLIANCE MANUFACTURERS: Make sure your company and products are listed free of charge in Light Reading's Security Appliances Directory, now in progress, by completing this questionnaire.