Security and Privacy in Industrial Cooperation

Protecting industrial control systems against cyberattacks is crucial to counter escalating threats to critical infrastructure. To this end, Industrial Intrusion Detection Systems (IIDSs) provide an easily retrofittable approach to uncover attacks quickly and before they can cause significant damage. Current research focuses either on maximizing automation, usually through heavy use of machine learning, or on expert systems that rely on detailed knowledge of the monitored systems. While the former hinders the interpretability of alarms, the latter is impractical in real deployments due to excessive manual work for each individual deployment. To bridge the gap between maximizing automation and leveraging expert knowledge, we introduce GeCo, a novel IIDS based on automatically derived comprehensible models of benign system behavior. GeCo leverages state-space models mined from historical process data to minimize manual effort for operators while maintaining high detection performance and generalizability across diverse industrial domains. Our evaluation against state-of-the-art IIDSs and datasets demonstrates GeCo’s superior performance while remaining comprehensible and performing on par with expert-derived rules. GeCo represents a critical step towards empowering operators with control over their cybersecurity toolset, thereby enhancing the protection of valuable physical processes in industrial control systems and critical infrastructures.

Cyberattacks have grown into a major risk for organizations, with common consequences being data theft, sabotage, and extortion. Since preventive measures do not suffice to repel attacks, timely detection of successful intruders is crucial to stop them from reaching their final goals. For this purpose, many organizations utilize Security Information and Event Management (SIEM) systems to centrally collect security-related events and scan them for attack indicators using expert-written detection rules. However, as we show by analyzing a set of widespread SIEM detection rules, adversaries can evade almost half of them easily, allowing them to perform common malicious actions within an enterprise network without being detected. To remedy these critical detection blind spots, we propose the idea of adaptive misuse detection, which utilizes machine learning to compare incoming events to SIEM rules on the one hand and known-benign events on the other hand to discover successful evasions. Based on this idea, we present AMIDES, an open-source proof-of-concept adaptive misuse detection system. Using four weeks of SIEM events from a large enterprise network and more than 500 hand-crafted evasions, we show that AMIDES successfully detects a majority of these evasions without any false alerts. In addition, AMIDES eases alert analysis by assessing which rules were evaded. Its computational efficiency qualifies AMIDES for real-world operation and hence enables organizations to significantly reduce detection blind spots with moderate effort.

Industrial control systems increasingly rely on middlebox functionality such as intrusion detection or in-network processing. However, traditional end-to-end security protocols interfere with the necessary access to in-flight data. While recent work on middlebox-aware end-to-end security protocols for the traditional Internet promises to address the dilemma between end-to-end security guarantees and middleboxes, the current state-of-the-art lacks critical features for industrial communication. Most importantly, industrial settings require fine-grained access control for middleboxes to truly operate in a least-privilege mode. Likewise, advanced applications even require that middleboxes can inject specific messages (e.g., emergency shutdowns). Meanwhile, industrial scenarios often expose tight latency and bandwidth constraints not found in the traditional Internet. As the current state-of-the-art misses critical features, we propose Middlebox-aware DTLS (Madtls), a middlebox-aware end-to-end security protocol specifically tailored to the needs of industrial networks. Madtls provides bit-level read and write access control of middleboxes to communicated data with minimal bandwidth and processing overhead, even on constrained hardware.

Industrial Intrusion Detection Systems (IIDSs) play a critical role in safeguarding Industrial Control Systems (ICSs) against targeted cyberattacks. Unsupervised anomaly detectors, capable of learning the expected behavior of physical processes, have proven effective in detecting even novel cyberattacks. While offering decent attack detection, these systems, however, still suffer from too many False-Positive Alarms (FPAs) that operators need to investigate, eventually leading to alarm fatigue. To address this issue, in this paper, we challenge the notion of relying on a single IIDS and explore the benefits of combining multiple IIDSs. To this end, we examine the concept of ensemble learning, where a collection of classifiers (IIDSs in our case) are combined to optimize attack detection and reduce FPAs. While training ensembles for supervised classifiers is relatively straightforward, retaining the unsupervised nature of IIDSs proves challenging. In that regard, novel time-aware ensemble methods that incorporate temporal correlations between alerts and transfer-learning to best utilize the scarce training data constitute viable solutions. By combining diverse IIDSs, the detection performance can be improved beyond the individual approaches with close to no FPAs, resulting in a promising path for strengthening ICS cybersecurity.