Sep. 25, 2002 UCLA computer scientists are developing a new tool that could prevent network operators from becoming unwitting hosts for a particularly harmful Internet-based threat: the distributed denial of service attack.
Distributed denial of service attacks — or DDoS — have interrupted service on Web sites like Yahoo, eBay, CNN and Amazon, and university campus networks nationwide have been shut down by them. A report released Sept. 18 by the federal President's Critical Infrastructure Protection Board has called upon universities to do more to protect their networks from Internet attacks.
Peter Reiher, adjunct associate professor of computer science at UCLA's Henry Samueli School of Engineering and Applied Science, claims a program called D-WARD could protect entire networks from being used as launch pads for destructive computer attacks. "D-WARD is a defense system that, when deployed on a network, will detect and stop attacks from being launched from that network," Reiher said.
In a DDoS attack, the victim's computer is hit by tens of thousands of data streams that quickly overload the network, denying service to legitimate users. A hacker initiates an attack by tapping into a number of computer networks, gaining control over thousands of machines, and causing each machine to generate traffic toward a particular target. Individual users may not even realize they have become unwilling participants — or "agents" — in a coordinated attack. Hackers use automated tools to compromise an agent computer or network within five seconds. Several thousand agents can be compromised in less than an hour.
D-WARD (an acronym for DDoS netWork Attack Recognition and Defense) protects networks from becoming would-be launch pads for an external hacker by stopping the crippling traffic streams before they have traveled very far. "Attacks can be stopped before they enter the Internet and blend with all the other traffic on its way to the victim," Reiher said. "Being close to the source can make it easier to trace an attack back to its origins."
The D-WARD module is installed on a network router, which serves as a gateway between a local network and the rest of the Internet. All traffic to and from the network passes through this router, and therefore, through D-WARD.
The module polices traffic to and from a network's IP addresses. If traffic flow begins to rise or fall outside normal patterns, the module looks more closely to determine if an attack is originating from that network. If it determines an attack is coming from a particular IP address or connection in the network, D-WARD moves quickly to cut the flow of traffic from that source, stopping an attack in its tracks.
D-WARD, which is still in its design stage, is the brainchild of UCLA graduate student Jelena Mirkovic. She has been working with Reiher on this project for almost two years, and feels that once D-WARD is ready for deployment, systems administrators everywhere will see the benefit of using this kind of defense system. "People who are security conscious will see that a single D-WARD system installed at the exit router would prevent DDoS attacks originating from the whole network," Mirkovic said.
A protected network also shields its owners from the social or financial implications of being used as an agent, Reiher said. "In the future those who do not take reasonable security measures to secure their system may be liable for damages inflicted by attacks coming from their machines."
Mirkovic explains that a defense system must be able to distinguish legitimate heavy traffic flow from an actual attack. "We're trying to catch scenarios where the victim is unable to respond to a large volume of information but continues to receive information anyway," she said.
Most Internet traffic, including e-mail, uses a protocol called TCP. This protocol is supposed to make sure data reliably gets from one user to another. It involves reverse messages, which are replies people receive after they send data to someone else — a file, an e‑mail or a request to view a Web page. "A user never sees these reverse messages," Reiher said, "but for each message you send, you should receive a response from the recipient saying he received your message. If the recipient stops responding, he may be experiencing a problem. While a legitimate user will wait until the congestion clears, an attacker won't stop."
When D-WARD detects that a network user is pumping out traffic without receiving reverse messages from the recipient, it reacts by reducing the amount of traffic that can be sent. This is called "rate-limiting."
"D-WARD is constantly classifying connections and flows and determining the appropriate rate-limits," Mirkovic said. "The module forwards packets belonging to good connections and analyzes suspicious packets more closely."
A nuanced approach to dealing with threats keeps the number of false-positives low. "We examine traffic closely before imposing rate-limits, and then continue to examine traffic once a rate-limit has been imposed to see if it is complying," Reiher said. "This is key to any successful automated approach."
An automated defense system like D-WARD can also assess threats quickly. System administrators spend a significant amount of their time watching for these kinds of things manually, searching records by hand, trying to recognize attack traffic. "That is very laborious," Reiher said. "If you can do it in an automated fashion, you can save all of that effort and still prevent your machines from being used for an attack on someone else."
Reiher and Mirkovic have been testing D-WARD on Linux-based software routers. They now plan to use new IXP programmable routers donated by Intel Corporation. The U.S. Defense Advanced Research Projects Agency (DARPA) has been funding the project since it began.
Reiher, who received both his Master's degree and Ph.D. at UCLA, has been on the faculty since 1992. He and Mirkovic plan to present their findings at an international conference on network protocols this November.
Other social bookmarking and sharing tools:
Note: Materials may be edited for content and length. For further information, please contact the source cited above.
Note: If no author is given, the source is cited instead.