Researchers have come up with host-based security software that blocks distributed denial-of-service attacks without swamping the memory and CPU of the host machines.The filtering, called identity-based privacy-protected access control (IPCAF), can also prevent session hijacking, dictionary attacks and man-in-the-middle attacks, say researchers at Auburn University in their paper, "Modeling and simulations for Identity-Based Privacy-Protected Access Control Filter (IPCAF) capability to resist massive denial of service attacks."

This new method is suggested as a replacement for IP-address filtering, which is sometimes used to block DDoS attacks but is problematic because IP addresses can be spoofed, says Chwan-Hwa "John" Wu, a professor of electrical and computer engineering at Auburn and lead author of the paper.

The method also greatly reduces the resources attacked machines have to expend in order to figure out whether requests are legitimate, he says.

Under IPCAF authorized users and the servers they try to reach receive a one-time user ID and password to authenticate to each other. After that they cooperate to generate pseudo IDs and packet-field values for each successive packet so packets get authenticated one at a time.

The receiving machines simply check the field value in each packet in order to decide whether to reject it. Only after the filter value checks out are more memory and CPU resources allocated to further process the packets, Wu says.

DDoS attacks can cripple access servers and disable resources needed by legitimate users, resulting in recent high-profile outages at Twitter and Bitbuket.

The major challenge to attacked machines is that they must commit memory and CPU resources to figuring out whether requests are legitimate. Separate appliances can sort out bad packets and proxy legitimate ones to the attacked machines, but that requires capital outlay for the devices and ongoing maintenance and management, he says.

By contrast, IPCAF runs on servers and client machines and does its work with negligible impact on performance of the machines involved, he says. For instance, the CPU on a machine running IPCAF and processing legitimate requests during testing was 10.21 per cent. That rose to 11.78 per cent when the same machine was under attack, Wu says.

He says machines using Pentium-class processors can filter packets in 6 nanosec using IPCAF, whereas the same machines would take a few milliseconds to make the same decision using public key infrastructure. That's about a million times slower with PKI. The significance is that the machines can get attack packets out of the way quickly before they start backing up and degrading response times to the point that users notice, Wu says.

To conserve processing power, IPCAF employs a lightweight hashing method -- hash-based message authentication code (HMAC) -- to generate the filter value it will use to authenticate each successive packet.

He says during lab tests, when a 10Gbps link to a server was filled with legitimate traffic and then with attack packets, network latency increased by 30 nanosec. "For humans, there is no difference," he says. Users don't sense the attack is underway, he says, and it remains possible for network security teams to trace the command and control center behind attacks.

Wu says the goal of the research is to create a commercial version of the software for use in business networks, but was uncertain when that might happen. He and his colleagues are still working on how their software might trace the source of attacks.