SECURE CLOCK SYNCHRONIZATION

FIELD OF THE INVENTION

The invention relates to industrial automation systems or Process Control systems, and in particular to Substation Automation systems, with an Ethernet based communication network.

BACKGROUND OF THE INVENTION

Process control or industrial automation systems are used extensively to protect, control and monitor industrial processes in industrial plants for e.g. manufacturing goods, transforming substances, or generating power, as well as to monitor and control extended primary systems like electric power, water or gas supply systems or telecommunication systems, including their respective substations. An industrial automation system generally has a large number of process controllers distributed in an industrial plant or over an extended primary system, and communicatively interconnected via a communication system.

Substations in high and medium- voltage power networks include primary devices such as electrical cables, lines, bus bars, switches, power transformers and instrument transformers, which are generally arranged in switch yards and/or bays. These primary devices are operated in an automated way via a Substation Automation (SA) system. The SA system comprises secondary devices, so-called Intelligent Electronic Devices (IED), responsible for protection, control and monitoring of the primary devices. The IEDs may be assigned to hierarchical levels, such as the station level, the bay level, and the process level, where the process level is separated from the bay level by a so-called process interface. The station level of the SA system includes an Operator Work Station (OWS) with a Human-Machine Interface (HMI) and a gateway to a Network Control Centre (NCC). IEDs on the bay level, which may also be referred to as bay units, in turn are connected to each other as well as to the IEDs on the station level via an inter-bay or station bus serving the purpose of exchanging commands and status information.

A communication standard for communication between the secondary devices of a substation has been introduced as part of the standard IEC 61850 entitled "communication networks and systems in substations". For non-time critical messages, IEC 61850-8-1 specifies the Manufacturing Message Specification (MMS, ISO/IEC 9506) protocol based on a reduced Open Systems Interconnection (OSI) protocol stack with the Transmission Control Protocol (TCP) and Internet Protocol (IP) in the transport and network layer, respectively, and Ethernet as physical media. For time-critical event-based messages, IEC 61850-8-1 specifies the Generic Object Oriented Substation Events (GOOSE) directly on the Ethernet link layer of the communication stack. For very fast periodically changing signals at the process level such as measured analogue voltages or currents IEC 61850-9-2 specifies the Sampled Value (SV) service, which like GOOSE builds directly on the Ethernet link layer. Hence, the standard defines a format to publish, as multicast messages on an industrial Ethernet, event-based messages and digitized measurement data from current or voltage sensors on the process level.

With the introduction of IEC 61850, precise time synchronization over Ethernet-based networks for secondary devices in process control or substation automation systems has become a concern. As a replacement of the classical Pulse-Per-Second PPS signal, IEC 61850 recommends the use of IEEE 1588 to achieve the degree of time synchronization required for critical data such as SV or trip signals. IEEE 1588 can run in two modes. In a one-step-clock mode, a master clock sends a synchronization message and at the same time timestamps the message and inserts the timestamp in the content of the same message. In a two-step-clock mode, the timestamp is not carried directly in the synchronization message but in a follow-up message.

A further prominent aspect in substation automation is the increased importance placed on cyber-security. While the protocols defined by IEC 61850, such as 8-1 and 9-2, are covered by IEC 62351-6 to define the required security mechanisms, IEEE 1588 remains unprotected. One of the problems to secure IEEE 1588 is the inability to secure the protocol when using a one-step-clock approach. A two-step-clock is trivial to secure, whether with a symmetric or asymmetric scheme, since the synchronization message is a non-sensitive message that is never modified at all. On the other hand, a secured one-step clock approach requires securing the synchronization message on the fly (while being forwarded), and thus is almost impossible (for an asymmetric scheme) or impossible (for a symmetric scheme or for a 1 Gbit/sec network) to implement.

The patent application EP 2148473 relates to mission-critical or highly-available applications based on a ring-type communication network with a plurality of switching nodes and operating with full duplex links. A sender node that is connected over a respective first and second port to the communication network transmits pairs of redundant frames. For each frame to be sent on the ring network, a source and a duplicate frame are transmitted in opposite directions, both frames being relayed by the other nodes of the ring network until they eventually return back to the originating sender node. As a consequence, network load is roughly doubled with respect to a conventional ring network, but the destination node will receive the data after a maximum transmission delay that equals to the longest possible path of the ring. In the fault-free state, the destination node thus receives two redundant frames with the same contents. The redundant frames can be identified according to a Parallel Redundancy Protocol PRP, hence only the earlier or first frame of the two frames is forwarded to the upper layer protocols and the later or second frame is discarded. As the synchronization message and the follow-up message of a two- step-clock approach may take different paths or directions in HSR, use of a one-step-clock is preferred.

The paper "Practical Application of 1588 Security", by Albert Treytl, Bernd Hirschler, IEEE International Symposium on Precise Clock Synchronization, September 2008, Ann Harbor, USA, proposes a way to implement a secure one-step-clock in which a static part of the synchronization message is hashed upfront. The timestamp is then produced and embedded in the message, following which the hash is completed rapidly on the remaining part of the message, i.e. the timestamp. The drawback of this approach is that it only allows a symmetric protection scheme which is less time consuming than the asymmetric one. Indeed, the operation is still done on the fly, i.e. hashing of the remaining part as well as signing, and is therefore time critical. Another drawback is the limitation of this approach to a 100 Mbit/sec network, since for a 1 Gbit/sec the remaining hashing operations will not be completed on time.

DESCRIPTION OF THE INVENTION

It is therefore an objective of the invention to secure a one-step IEEE 1588 clock using either a symmetric or asymmetric protection scheme. This objective is achieved by a method of synchronizing clocks and a master clock device according to the independent claims. Preferred embodiments are evident from the dependent patent claims, wherein the claim dependency indicated shall not be construed as excluding further meaningful claim combinations.

According to the invention, clocks of mission-critical or highly-available devices in industrial automation systems connected to a communication network are synchronized by sending, by a master clock, a synchronization message, in particular a single message of the one-step-clock type according to IEEE 1588, including a time stamp, and by receiving and evaluating, by a slave clock, the synchronization message. A synchronization component or module of the master clock prepares, or composes, prior to a projected send time t _S end, a synchronization message including a time stamp of the projected send time, and secures the synchronization message still in advance of the projected send time. Securing the synchronization message takes place by suitable cryptographic means allowing at least for authentication of the time stamp at a receiving slave clock, e.g. by calculating and signing a checksum or hash of the synchronization message. At the projected send time, the secured synchronization message is transmitted.

In an advantageous embodiment of the invention, the secured synchronization message is stored in a dedicated wait component of a transmitter of the master clock device prior to being sent at t _sen d.

In another advantageous embodiment of the invention, a Low Priority Queue (LPQ) of the transmitter is blocked and sending of non-synchronization messages from the LPQ is disabled during a blocking interval prior to t _sen d. A conservative blocking interval corresponding to the longest message expected in the LPQ ensures that no message is in the process of being transmitted at t _sen d. In a more sophisticated variant, the length of a message from the LPQ is checked prior to sending in order to ascertain completion of its transmission prior to t _sen d. Hence, the synchronization messages will always be transmitted without further delay or jitter due to ongoing transmission of low priority messages.

Preferably, the clock devices are arranged as switching nodes in a ring-type communication network operating with full duplex links, wherein a sender node that is connected over a respective first and second port to the communication network transmits pairs of redundant frames. For each synchronization message to be sent on the ring network, a source and a duplicate synchronization message are transmitted in opposite directions, both messages being relayed by the other nodes of the ring network until they eventually return back to the originating sender node. Such redundant communication network topology is advantageously employed in industrial automation systems or Process Control systems, in particular in Substation Automation systems for substations in high and medium- voltage electric power networks.

BRIEF DESCRIPTION OF THE DRAWINGS

The subject matter of the invention will be explained in more detail in the following text with reference to preferred exemplary embodiments which are illustrated in the attached drawings, in which:

Fig. l depicts a sequence of operations for a secure one-step-clock implementation outside the scope of the present invention;

Fig.2 depicts a modified sequence of operations for a secure one-step-clock; and

Fig.3 and Fig.4 depict corresponding system architectures. DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

Fig. l depicts an exemplary but rather impractical sequence of operations required for a straight-forward implementation of a secure one-step IEEE 1588 clock. The main problem associated therewith is the inability to compute all the required operations on the fly. In particular, such a clock would have to start sending the first byte of the synchronization message SYNC, then time stamp the sending operation, insert the time stamp in the message, and finally hash and sign the message with the shortest possible time delay.

Fig.2 depicts an exemplary sequence of operations of a secure IEEE 1588 one-step clock preparing synchronization messages in advance. This invention proposes to prepare the synchronization message SYNC for a future time stamp, i.e. to produce and embed the time stamp for a future time t _sen d = t _pre p + At, where the time t _prep designates the time at which the preparation of the synchronization message starts and its advanced time stamp is determined. By carefully choosing the preparation interval or advance delay At, sufficient time to perform all necessary operations prior to sending the secured synchronization at the projected send time t _sen d is available.

Fig.3 shows the corresponding system architecture of a typical one-step clock implementation. It is the ultimate responsibility of the IEEE 1588 low-level stack to ensure that the synchronization message is sent on time, and the proposed improvement is implemented in hardware to achieve required timing precision. The hardware level, or low- level stack, comprises a dedicated Integrated Circuit (IC) chip or Field Programmable Gate Array (FPGA), preferably as part of the Network Interface Card (NIC) 10 of the device, and executes all time-critical aspects of the synchronization procedure, i.e. time stamping IEEE 1588 SYNC messages when they are received and sent. The logical gates of the Time Stamping Unit (TSU) 11 are coded by way of a low programming language such as VHDL. On a higher abstraction level, all non-time critical aspects are implemented within the regular IEEE 1588 stack 14, either on the same dedicated chip or on the CPU 13 of the device, the latter being connected to the TSU via exemplary communication bus (e.g. PCI) 12. In this case, the IEEE 1588 stack runs on the CPU as a software implementation, while only the time stamp operation is achieved at the hardware level. The proposed invention does not imply any modification on the software stack but only on the logic implementing the sequence of steps according to Fig.2, i.e. when sending a synchronization message.

Fig.4 finally depicts the detailed architecture of a low level stack supporting the secure IEEE 1588 one-step-clock formed in advance. The various components should be understood from an architecture point of view (i.e. in terms of software engineering) as entities performing a function and having inputs and outputs. As mentioned above, from an implementation point of view, a component can implemented as a VHDL component (i.e. dedicated gates with some memory) if implemented in FPGA, but also a C function or Java object if implemented in software.

The receiver logic illustrated in Fig.4 remains unmodified compared to a normal implementation of IEEE 1588 with hardware support. Its purpose is to decode the messages received from the network, and to detect the arrival of a synchronization message requesting a time stamping operation to be performed by the TSU. Similarly, the transmit logic detects the presence of an outbound synchronization message, originating e.g. from the CPU of the device, and likewise requesting a time stamping operation. The sequence of operations represented in Fig.2 is implemented in the components Sync 20 and Wait 21. The former is responsible for preparing, including receiving the timestamp from the TSU, the fully secured synchronization message, while the latter withholds sending the message until the projected send time t _sen d. The transmission port TX contains two queues, a High Priority Queue (HPQ) 21 dedicated to IEEE 1588 secure synchronization messages and a Low Priority Queue (LPQ) 22 for any non-sync messages. The HPQ 21 has the highest priority, meaning that whenever a message is placed in the queue, the message will be transmitted without further delay. Nevertheless, in case the transmitter has just started sending a message from the LPQ 22, the sending of the synchronization message is delayed by [(sizeof(max_length_ethernet jacket) + interframe_gap) / network _sp >eed] ^' . This leads to a maximum delay of (1526x8+ 12x8)/(100Mbits/sec) = 12.6 μ εε for a 802.3 MAC frame on a 100 Mbit/sec network. To prevent this additional uncontrolled jitter, the Wait component 21 has to block or disable the sending of any frame from the LPQ in a blocking interval Abi _oc k of 12 prior to t _sen d.

The choice of the preparation interval At depends on many parameters such as the hardware used, the security scheme, the speed of the network, etc. For example, the VHDL implementation of a specific hash function known as AES (Advanced Encryption Standard) requires between 50 (high performance) and 106 (low performance) cycles of 20 ns (50 MHz) for hashing 6 bytes, resulting in a delay of about 33 μβεΰ or 70 μβεΰ for a typical 200 bytes long synchronization frame. In this case, At must be at least 33 μβεΰ or 70 μβεΰ plus the required delay to insert the time stamp in the synchronization message.