The Siemens S7 protocol is commonly used in SCADA systems for communications between a Human Machine Interface (HMI) and the Programmable Logic Controllers (PLCs). This paper presents a model-based Intrusion Detection Systems (IDS) designed for S7 networks. The approach is based on the key observation that S7 traffic to and from a specific PLC is highly periodic; as a result, each HMI-PLC channel can be modeled using its own unique Deterministic Finite Automaton (DFA). The resulting DFA-based IDS is very sensitive and is able to flag anomalies such as a message appearing out of its position in the normal sequence or a message referring to a single unexpected bit. The intrusion detection approach was evaluated on traffic from two production systems. Despite its high sensitivity, the system had a very low false positive rate -over 99.82% of the traffic was identified as normal.

SCADA protocols for Industrial Control Systems (ICS) are vulnerable to network attacks such as session hijacking. Hence, research focuses on network anomaly detection based on meta-data (message sizes, timing, command sequence), or on the state values of the physical process. In this work we present a class of semantic network-based attacks against SCADA systems that are undetectable by the above mentioned anomaly detection. After hijacking the communication channels between the Human Machine Interface (HMI) and Programmable Logic Controllers (PLCs), our attacks cause the HMI to present a fake view of the industrial process, deceiving the human operator into taking manual actions. Our most advanced attack also manipulates the messages generated by the operator's actions, reversing their semantic meaning while causing the HMI to present a view that is consistent with the attempted human actions. The attacks are totaly stealthy because the message sizes and timing, the command sequences, and the data values of the ICS's state all remain legitimate. We implemented and tested several attack scenarios in the test lab of our local electric company, against a real HMI and real PLCs, separated by a commercial-grade firewall. We developed a real-time security assessment tool, that can simultaneously manipulate the communication to multiple PLCs and cause the HMI to display a coherent system-wide fake view. Our tool is configured with message-manipulating rules written in an ICS Attack Markup Language (IAML) we designed, which may be of independent interest. Our semantic attacks all successfully fooled the operator and brought the system to states of blackout and possible equipment damage.

No abstract

Traffic of Industrial Control System (ICS) between the Human Machine Interface (HMI) and the Programmable Logic Controller (PLC) is known to be highly periodic. However, it is sometimes multiplexed, due to asynchronous scheduling. Modeling the network traffic patterns of multiplexed ICS streams using Deterministic Finite Automata (DFA) for anomaly detection typically produces a very large DFA, and a high false-alarm rate. In this paper we introduce a new modeling approach that addresses this gap. Our Statechart DFA modeling includes multiple DFAs, one per cyclic pattern, together with a DFA-selector that de-multiplexes the incoming traffic into sub-channels and sends them to their respective DFAs. We demonstrate how to automatically construct the statechart from a captured traffic stream. Our unsupervised learning algorithms first build a Discrete-Time Markov Chain (DTMC) from the stream. Next we split the symbols into sets, one per multiplexed cycle, based on symbol frequencies and node degrees in the DTMC graph. Then we create a sub-graph for each cycle, and extract Euler cycles for each sub-graph. The final statechart is comprised of one DFA per Euler cycle. The algorithms allow for non-unique symbols, that appear in more than one cycle, and also for symbols that appear more than once in a cycle.We evaluated our solution on traces from a production ICS using the Siemens S7-0x72 protocol. We also stress-tested our algorithms on a collection of synthetically-generated traces that simulated multiplexed ICS traces with varying levels of symbol uniqueness and time overlap. The algorithms were able to split the symbols into sets with 99.6% accuracy. The resulting statechart modeled the traces with a median false-alarm rate of as low as 0.483%. In all but the most extreme scenarios the Statechart model drastically reduced both the false-alarm rate and the learned model size in comparison with the naive single-DFA model.

No abstract