Sinkhole (computers)

From Citizendium
Revision as of 08:55, 12 June 2009 by imported>Howard C. Berkowitz
Jump to navigation Jump to search
This article is developing and not approved.
Main Article
Discussion
Related Articles  [?]
Bibliography  [?]
External Links  [?]
Citable Version  [?]
 
This editable Main Article is under development and subject to a disclaimer.

In computer network and server security, a sinkhole is a network element to which suspect or dangerous traffic is diverted. Sinkhole techniques move denial of service traffic away from the network as a hole, possibly disabling a target server but protecting the overall network infrastructure.

Sinkholes often contain extra analysis tools that can variously confirm suspect traffic is indeed an attack, assist in or even automatically create defenses, and help trace an attack back to the source.

Large networks, as with Internet Service Providers, often have multiple sinkholes distributed to multiple locations, so diverted traffic can be processed close to the point that it enters the network, rather than creating an undesirable load if it has to reach a distant forensic analysis point. Since the sinkholes usually are identical, the anycasting technique is very useful to advertise their presence: it can be easy to administer, and is inherently load-balancing and fault-tolerant.

Sinkhole implementation can differ when the emphasis is protecting routers rather than servers, although there are some common techniques. For router protection, it must be possible to divert attack traffic directed at the usual production address of the router interface, but still have either an additional, internal-only address on the interface, or another means of communicating with the router (e.g., a physically separate interface) that can be used to reconfigure the router. This is less often required on servers, which are most often protected by router configuration, but still may be appropriate if defensive code needs to be inserted into the router, yet not diverted away from it by the anycast process. [1] When using anycast with routers, it is best to use a routing protocol, such as the Border Gateway Protocol or Open Shortest Path First, that has an explicit router identifier that can disambiguate different announcements of the same address.

References

  1. John Kristoff (January 2, 2004), Anycast Routing, Anycast Addressing on the Internet