Scaling intrusion prevention systems for 10G, 40G and beyond
The Internet is still growing and, as we make the transition from Intrusion Detection Systems (IDS) to Intrusion Prevention Systems (IPS) – driven in part by the fact that IPS technology can scale to meet users’ needs in the future. Against this backdrop, this article takes a look at the distinctive features of IPS and how the performance of these systems can be increased to keep up with the relentless development on the bandwidth front.
Whilst many IT security professionals regard IPS to be a natural extension of IDS technology, the real answer is that IPS is actually another type of access control mechanism, rather than purely a sister IT security platform to IDS.
It may surprise you to know that the term IPS is actually a lot younger than IDS, and is a colloquial term first used by Andrew Plato, a technical consultant with a major IT security vendor that, way back in the late 1990s, developed the industry’s first IDS platform. In its purest form, an IPS makes a number of access control decisions based on the content of the application, rather than taking a traditional firewall approach of monitoring IP addresses, ports and other connective links.
Back in 1998, Andrew Plato opined that a good IPS should feature a sophisticated analytical engine, but one that generates as few false positives as possible. Provided this is the case, he said at the time, then a good IPS has a number of advantages over IDS, since it can sit in line with an IP traffic flow and analyze the data stream in real time. In addition, most modern IPS solutions also have the ability to analyze Layer 7 protocols such as FTP, HTTP and SMTP, and make decisions on whether to allow – or quarantine – the IP packets as required, even if the data is encrypted.
But are today’s IPS platforms up to the task of scanning IP traffic at the high speeds needed in a modern IT environment? The problem facing IT professionals is that, with the Internet growing at between 40 and 60 per cent a year (source: Atlas Internet Observatory) – and against the backdrop of a mobile data explosion – it’s important that IPS technology can keep up with this data bandwidth growth and not become the bottleneck in the network.
It’s also becoming clear that, on a typical network of today, users are placing a very load on each port of a multi-10G port system and, whilst there are IPS products available that are capable of supporting a multiple 10 Gbps port topology, providing continuous 10 Gbps throughput on these ports is a something of a challenge. The most worrying part of this development is how IPS platforms can be scaled to meet the needs of 40G and 100G IPS technologies, which are set be introduced to the IT/networking mix in the next few years.
Until a few years ago, it could be argued that IPS platforms were up to the task, especially since most IPS platforms adopted a core five-stage real-time analysis process that steps through a number of stages as various IT threats are encountered when monitoring an organization’s data streams that flow both in and out of the IT resource.
The first stage is to bandwidth throttle any suspicious IP traffic to give the organization’s IT security software a chance to analyse the data stream – let’s take the example of an email message stream – and deal with suspect messages and/or attachments in real time. If the data is found to be suspect, but does not conform to known infection signatures, then the second stage is for the message’s header to be analyzed and, if an infection etc., is found, the data can be quarantined.
The third stage in the analysis process involves performing user management and address validation, typically by applying a number of automated checks to verify whether the message comes from a source previously known to be dangerous. The fourth stage involves applying an anti-malware and anti-hacking analysis engine for anything suspicious that has passed the first three analysis stages but does not pass muster. The fifth stage, typically involves using the analysis engine to weed out anything that still looks suspicious for later, manual, analysis by the IT security staff concerned.
However, the increasing sophistication of malware, together with the recursive and obfuscated coding approach taken by an increasingly criminal hacking fraternity – and, of course, the higher network speeds seen on today’s networking systems – means that IPS systems are under pressure to keep up, both in terms of handling the amount of data, but also in having the raw horsepower to run more sophisticated algorithms.
Raising the security game
Businesses are not just increasing their IT security budgets, however, they are also raising their game when it comes to security strategies. Given this scenario, the key challenge now is scaling these systems to keep up with the increasing bandwidth generated by richer content in emails and on Web sites, more video and teleconferencing and the transition to cloud computing already taking place.
The important thing to realize is that all these innovative services provide a new and high-speed avenue of attack for hackers. And because of this, network security systems need to react in real-time to contain the problem. To keep up with these high-speed, real-time demands, the traditional approach of network security appliance vendors has been to invest in the development of customized, proprietary hardware. However, a new approach is emerging where off-the-shelf standard PC server hardware is being used negating the need for hardware development.
A Napatech poll revealed that the majority of network security appliances being used are still based on proprietary hardware, but for every 3 proprietary systems, there are now 2 systems based on standard PC server hardware. In the past, PC servers have not been powerful enough to meet the demands of security applications like IPS, but the latest generation of PC servers provide significant processing power and a strong roadmap of increased performance to come.
Our researchers have discovered it has now become more economical to build network security appliances based on standard PC server hardware using real-time network analysis adapters to ensure high levels of performance. The most compelling reasons for considering this approach is the ability to scale performance.
CPU makers such as Intel and AMD are increasing the performance of their chips by as much as 50 per cent on an annual basis. Can the vendors of proprietary network security appliances keep up with this kind of performance roadmap? Does it even make sense to try? The time has come to reconsider.