Active fuzzing for testing and securing cyber-physical systems

Abstract

Cyber-physical systems (CPSs) in critical infrastructure face a pervasive threat from attackers, motivating research into a variety of countermeasures for securing them. Assessing the effectiveness of these countermeasures is challenging, however, as realistic benchmarks of attacks are difficult to manually construct, blindly testing is ineffective due to the enormous search spaces and resource requirements, and intelligent fuzzing approaches require impractical amounts of data and network access. In this work, we propose active fuzzing, an automatic approach for finding test suites of packetlevel CPS network attacks, targeting scenarios in which attackers can observe sensors and manipulate packets, but have no existing knowledge about the payload encodings. Our approach learns regression models for predicting sensor values that will result from sampled network packets, and uses these predictions to guide a search for payload manipulations (i.e. bit flips) most likely to drive the CPS into an unsafe state. Key to our solution is the use of online active learning, which iteratively updates the models by sampling payloads that are estimated to maximally improve them. We evaluate the efficacy of active fuzzing by implementing it for a water purification plant testbed, finding it can automatically discover a test suite of flow, pressure, and over/underflow attacks, all with substantially less time, data, and network access than the most comparable approach. Finally, we demonstrate that our prediction models can also be utilised as countermeasures themselves, implementing them as anomaly detectors and early warning systems.

Adversarial Attacks and Mitigation for Anomaly Detectors of Cyber-Physical Systems

“…Our study is based on the SWaT dataset [18] and WADI dataset [35], which are publicly available [53] and have been used in multiple projects [12,20,21,54,55,56,57]. The SWaT dataset records the system state of 26 sensor values and 25 actuators (in total 51 features) every second, and WADI records 70 sensors and 51 actuators (in total 121 features) every second.…”

“…The effectiveness of an anomaly detector can be assessed by subjecting it to a test suite of attacks, and observing whether it can correctly identify the anomalous behaviour. These tests can be derived from benchmarks [18], hackathons [19], or tools such as fuzzers [20,21], and typically involve manipulating or spoofing the network packets exchanged between CPS components. While studies have shown that neural network-based detectors are effective at detecting these conventional types of attacks [18,22,23,24,25,26,27], less consideration has been given to testing their effectiveness at detecting adversarial attacks, in which attackers have knowledge of the model itself and craft noise (or perturbations 1 ) that is specifically designed to cause data to be misclassified.…”

CPFuzz: Combining Fuzzing and Falsification of Cyber-Physical Systems

“…This paper searches the actuator configuration in a bit vector under the assumption that the states of actuators are discrete. In that vein, their following research [20] used online active learning to guide a search for network packet payload, which encodes actuator commands to drive the CPS into an unsafe state. In our paper, there are a couple of differences to notice: none of this line of work exploited the program coverage information, which may provide useful information for optimization, while our approach is based on coverage guided fuzzing technique, which has many success stories in largescale software [21]- [23].…”

“…If the position, velocity and acceleration of the pendulum enter a certain area, the control system fails. We try to falsify the property that the system never enter the following region of the state-space 20,20]. Sampled Polarity Integrator System [11].…”

Code integrity attestation for PLCs using black box neural network predictions

“…Compromising any one of these components or PLCs can potentially allow an attacker to manipulate the system into a damaging physical state. This has motivated a huge variety of research into defending and assessing CPSs, spanning techniques based on anomaly detection [7,12,17,22,28,32,39,40,44,45,47,49,51,54,57,60,65,66], fingerprinting [13,35,43,50,79], invariant-based monitoring [8,9,11,20,24,25,29,73,81], trusted execution environments [70], and fuzzing [26,27,78].…”