Changes in the network security landscape

Paul Boughton

John Browett considers the current position and looks at how to mitigate potential threats

Until a few years ago network security focused on enterprise systems and relied on firewalls to prevent viruses from arriving electronically. Now there is increasing realisation that the production control systems need to be protected too and that viruses can be uploaded by many means. Industries as diverse as manufacturing, processing, refining, power generation, and food and drink processing are all considering ways to improve the security of their plant and control systems.

As Ethernet becomes the de facto industrial network of choice, installers and users can be seduced by the advantages the technology brings and give insufficient consideration to network security. However, given some recent high profile situations there is now a growing awareness that there is the very real possibility that networks can be compromised by both the electronic importation of a virus and from within, either accidentally or maliciously.

The problems of hacking from within a company are as much a personnel security issue as a general network security issue. Security considerations need to consider both deliberate acts of sabotage and the possibility of personnel making an unintended mistake.

Further, companies are increasingly adopting systems that allow remote access to plants – and reaping great benefits. However, monitoring processes typically use standard web browsers, which open the system up to the possibility of abuse of the network by third parties.

It is now well understood that SCADA (supervisory control and data acquisition) and other plant floor control systems have weaknesses and vulnerabilities when it comes to security. Therefore many companies are reassessing their traditional methods for moving information around between the plant/asset and the enterprise level.

The point of attack can be at the enterprise system level, plant control level or even the individual field device level. Top end attacks have previously been the main concern, with the result that very sophisticated security measures are available. Vulnerable field devices are relatively easy to protect with local measures.

However in the ‘middle ground’ of plant control we frequently see PC-based control systems with little or no security measures in place. There are even cases of some technologies still being utilised despite known vulnerabilities.

Security problems at this level and at plant floor device level are exacerbated by the fact that there is often limited collaboration between a company’s IT department and the control engineering departments. In addition, within the control and engineering community, there is not always adequate recognition of the automation system security threats and liabilities. In particular, the business case for automation system security is not established, and there is limited understanding of the automation system risk factors.

The drive towards open network technologies generally, and towards Ethernet in particular, as a means of giving companies the freedom they want to choose best-of-breed control technologies has exacerbated the security threat. Users want standardisation, flexibility and choice, and this has been delivered through standardised open protocols. The trade-off, though, which is only just coming to be realised, is that these open protocols are less robust and more susceptible to attack. By contrast, the old proprietary networks were highly robust by virtue of their non-standardisation, but they were far less flexible and they ultimately limited product choice.

Looking then at what the ideal industrial network would offer, we can build up a wish list that offers the robustness of the old combined with the flexibility of the new. This wish list might include common cabling, standard connectors, open standards, ease of configuration, flexibility, highest possible security, and reduced susceptibility to attack.

In looking at how we might be able to adapt industrial Ethernet to meet the requirements of this wish list, it is worth revisiting our definition of Ethernet, because nowhere in networking parlance has a single word been so misused as an umbrella term for so many disparate standards, technologies and applications. And the best place to start for that is with the OSI seven layer model itself.

Layer 1, the Physical Layer, defines all the electrical and physical specifications for devices. In particular, it defines the relationship between a device and the physical medium. Layer 2 is the Data Link Layer, providing the functional and procedural means to transfer data between network entities and to detect and possibly correct errors that may occur in the Physical Layer. It is here that Ethernet is defined as a network protocol under the IEEE 802.3 standard.

Over the years, Ethernet has become synonymous with the TCP/IP suite, but one does not necessarily imply the other. IP is defined under the Network Layer (Layer 3) of the OSI model. This Layer provides the functional and procedural means of transferring variable length data sequences from a source to a destination via one or more networks. The Transport Layer (Layer 4) provides transparent transfer of data between end users, and defines the likes of TCP and UDP.

The Session Layer (Layer 5) controls the connections between computers, whilst the Presentation Layer (Layer 6) transforms the data to provide a standard interface for the Application Layer (Layer 7) at the top of model. It is here that you find typical applications such as FTP, HTTP, RTP, SMTP, SNMP and others. In short, when it comes to operating as a communications architecture in industrial networks, Ethernet is capable of very little without the layers that sit above it.

Not all industrial Ethernet offerings implement the Ethernet stack in the same way. Within the Application Layer the different industrial Ethernet organisations implement their own kernels and protocols which define much of the functional benefits of their technologies. From a security point of view, though, what is really of interest are the more vulnerable lower layers.

Under the seven layer model, all it takes is for one layer to fall to an attack before the whole communications system is compromised – potentially without the other layers even being aware that there is a problem. Security is only as strong as the weakest link.

There are a number of discrete security products available, and these work well, but one of the biggest problems in the industrial arena lies in implementing tightly integrated security systems without incurring excessive costs and without imposing a level of complexity that makes the system difficult to maintain and support. Further, standard commercially available security solutions are rarely up to the rigours of life in challenging industrial environments.

In terms of network technology, much work has been done to make Layer 2 more secure, but in classic implementations of industrial Ethernet little has been done to address weaknesses in the Network Layer (Layer 3) and the Transport Layer (Layer 4). Like the office Ethernet implementation, the vast majority of industrial Ethernet technologies are still built around IP within Layer 3 and TCP/UCP within Layer 4.

Most industrial Ethernet network installations implement perimeter security (firewall services) at points where they connect to other networks to provide protection at these vulnerable layers. Firewalls filter on source and destination IP addresses and protocol port numbers (for example TCP and UDP ports) to further restrict the traffic permitted to enter an Ethernet network. Packet filtering may be implemented even among known network communities, and in some cases filtering deals with very specific device addresses and application ports to provide a layer of access security unique to an attached device and application. Despite this however, in classic industrial Ethernet implementations, Layer 3 and Layer 4 are still highly vulnerable to attack.

CC-Link IE, however, is different. CC-Link IE (Control and Communication Link Industrial Ethernet) was developed by CLPA as the first completely integrated gigabit Ethernet network for industrial automation, defining the new threshold for open standards for Industrial Ethernet. CC-Link IE combines the best of many existing technologies and applies them to an optical or copper based industrial network system with a redundant architecture that enables extremely high-speed and reliable data transfer between field devices and other controllers via Ethernet links. The signalling rate of 1Gbps will redefine the users’ expectations and systems capabilities; it being more than enough to cater for the real-time communications requirement of today’s manufacturing industries.

There are variants of CC-Link IE to address control requirements at all levels of the automation network. At controller level, there is CC-Link IE Control. At device level, there is CC-Link IE Field and CC-Link IE Motion. And of course there is tight integration with the CC-Link fieldbus.

Most importantly, CC-Link IE differs from conventional implementations by defining an open 'Real-Time Protocol' within the stack layers. By taking this approach to implementing these layers within the Ethernet stack, CC-Link IE realises the benefits of our network technology wish list. It uses standard Ethernet connectors, it is easy to configure and it is highly robust. It is also an open standard, so users still have that freedom of choice in the selection of best-of-breed component technologies. But most importantly from a security point of view, it inherently offers the highest possible security and is therefore less susceptible to attack. These are significant advantages over alternative industrial Ethernet implementations. The key distinguishing factor is an open, but controlled knowledge base for the network technology. Hence while bona-fide companies can implement the technology on an open basis, it will be harder for the 'bad guys' to infiltrate.

Security requirements for industrial Ethernet networks are continuing to evolve, with sophisticated requirements increasingly migrating from Enterprise networks to process control and other industrial environments. Wherever there are network installations, companies need to look at the probability of attacks to the network, and the risk associated with any attack. In every case, as security becomes more important, companies must look at ways to mitigate the risk, reduce the risk or eliminate the risk as appropriate within each branch of the network topology. With its open standards approach combined with proprietary communications technology, the CC-Link IE implementation of industrial Ethernet represents an extremely attractive option in the drive to maximise and optimise network security.

John Browett is General Manager, CC-Link Partners Association-Europe