Internet breakdown

The internet is constructed around the Open Systems Interconnection (OSI) seven-layer model: at the bottom level are the wires and cables that connect computer to computer, a complex network which would be difficult to knock out completely.

When an under-sea internet cable was damaged by a ship’s anchor last year, disrupting web traffic to and from India, Pakistan, Sri Lanka and the Middle East, service became erratic but never stopped. Messages were simply rerouted through other parts of the network. Nevertheless, it took a couple of weeks to restore the damaged cable.

At the top of the OSI stack are the Domain Name System (DNS) servers, 13 large computers complete with back-up devices, all of them securely installed in “hardened” sites in the US.

The DNS is the “telephone directory” of the internet, responsible for keeping track of the web as it grows and for translating human-friendly internet addresses into numerical symbols the computers can understand.

These computers are almost certainly the weakest link. Again, while unlikely, it would be possible to lose all 13 as a result of terrorist activity or natural calamity.

Indeed, the closest the public internet has come to catastrophic failure was in January 2001 when large-scale “denial-of-service” (DoS) attacks – which typically involve bombarding a target server with spurious requests for information to slow it down or stop it – took 10 of the 13 DNS servers offline.

The internet itself, however, continued to function because caching servers – computers, often owned by internet service providers (ISPs) or large commercial organisations, which operate a subset of the DNS function – were unaffected. Caching servers can function autonomously for some time. Eventually, however, they have to refer to the main DNS servers for system upgrades, which would cause disruption.

As a result of the 2001 attacks, “anycast” load-balancing technology, which routes traffic to the nearest available server, was deployed. This proved effective against a second DoS attack in 2007. Six DNS servers were targeted but only three – where anycast technology had yet to be installed – failed.

The consensus is that it is unlikely that anyone could bring down the entire internet. But there are vulnerabilities which could be exploited by the ill intentioned at every level of the OSI model. For obvious reasons these weaknesses are not made public.

Individual companies are continually at risk from hackers. Last month, the blogging service Twitter and the social networking site Facebook were both hit by DoS attacks which left them out of action for an hour or more.

And nation states are at significant risk from cyber warfare. Two years ago, computers in Estonia, a country which depends largely on the internet for government communication, was subjected to a sustained DoS attack which came close to bringing the country to its knees.

The targets included the country’s foreign and defence ministries, leading newspapers and banks and the attacks continued for two weeks. Some suspect that Russian interests were responsible, in retaliation for the moving of a Soviet war memorial in Tallinn but culpability has never been proved or admitted.

If an entire country could be paralysed by internet failure, the consequences for individual businesses would be proportionately more severe.

A survey, carried out last year by the consultancy Quocirca on behalf of the ISP Easynet Connect, indicated that almost three quarters of small and medium sized businesses in the UK thought they would not cope for more than 24 hours without the internet.

The same survey showed that four out of 10 use it for remote back-up and disaster recovery. This picture is likely to be true for most, perhaps all, developed economies.

However, the overall picture is so complicated it is impossible to predict the consequences of failure. Much would turn on individual and collective responses. If these are rational and controlled, restoring the status quo would be much easier than if there were widespread panic.

Received wisdom in the intelligence community says: “The population is only four missed meals away from civil unrest.” But according to research published last month by the UK Office of Communications, 50 per cent of the country’s consumers would rather cut back on eating out, home improvements and holidays than give up communications services.

What should businesses do to protect themselves against the possibility of an internet crash?

First, and most important, they should review their business processes to identify points of possible dependence on the public internet. The big banks and main stock exchanges, for example, are likely to use dedicated lines, rather than the public internet and would not be affected directly by a crash.

But smaller financial organisations may not be in this situation, and large manufacturers, dependent on just-in-time provisioning, could be brought to a standstill by problems at their suppliers caused by lack of internet access.

Security specialists point out that, while the internet itself is extraordinarily resilient, it is vulnerable to the loss of its lifeblood – electricity supply. The effects of power cuts would be regional rather than global but extremely disruptive nonetheless.

Of course, in that case, loss of the internet would not be the prime consideration. Without electricity, towns and cities would quickly run short of water, which would put an inability to “tweet” in the shade.

Uninterruptible power supplies which continue to provide electricity in the event of grid failure are one answer.

But even they would not be effective in the event of a “coronal mass ejection” a ball of superheated ionised gas thrown off from the surface of the sun during a violent solar storm. If the gas were to hit our atmosphere it would create electromagnetic changes that would melt the wiring in electrical transformers, wiping out power grids across the globe.

And the likelihood of such a calamity? On balance, it is greater than the simultaneous loss of all 13 DNS servers.