Date of Award

August 2019

Degree Type


Degree Name

Doctor of Philosophy (PhD)


Mechanical and Aerospace Engineering


Young B. Moon


Alert Correlation, Cyber-Manufacturing System, Cyber-Physical System, Intrusion Detection, Security

Subject Categories



In the vision of Cyber-Manufacturing System (CMS) , the physical components such as products, machines, and tools are connected, identifiable and can communicate via the industrial network and the Internet. This integration of connectivity enables manufacturing systems access to computational resources, such as cloud computing, digital twin, and blockchain. The connected manufacturing systems are expected to be more efficient, sustainable and cost-effective.

However, the extensive connectivity also increases the vulnerability of physical components. The attack surface of a connected manufacturing environment is greatly enlarged. Machines, products and tools could be targeted by cyber-physical attacks via the network. Among many emerging security concerns, this research focuses on the intrusion detection of cyber-physical attacks.

The Intrusion Detection System (IDS) is used to monitor cyber-attacks in the computer security domain. For cyber-physical attacks, however, there is limited work. Currently, the IDS cannot effectively address cyber-physical attacks in manufacturing system: (i) the IDS takes time to reveal true alarms, sometimes over months; (ii) manufacturing production life-cycle is shorter than the detection period, which can cause physical consequences such as defective products and equipment damage; (iii) the increasing complexity of network will also make the detection period even longer. This gap leaves the cyber-physical attacks in manufacturing to cause issues like over-wearing, breakage, defects or any other changes that the original design didn’t intend.

A review on the history of cyber-physical attacks, and available detection methods are presented. The detection methods are reviewed in terms of intrusion detection algorithms, and alert correlation methods. The attacks are further broken down into a taxonomy covering four dimensions with over thirty attack scenarios to comprehensively study and simulate cyber-physical attacks.

A new intrusion detection and correlation method was proposed to address the cyber-physical attacks in CMS. The detection method incorporates IDS software in cyber domain and machine learning analysis in physical domain. The correlation relies on a new similarity-based cyber-physical alert correlation method. Four experimental case studies were used to validate the proposed method. Each case study focused on different aspects of correlation method performance. The experiments were conducted on a security-oriented manufacturing testbed established for this research at Syracuse University.

The results showed the proposed intrusion detection and alert correlation method can effectively disclose unknown attack, known attack and attack interference that causes false alarms. In case study one, the alarm reduction rate reached 99.1%, with improvement of detection accuracy from 49.6% to 100%. The case studies also proved the proposed method can mitigate false alarms, detect attacks on multiple machines, and attacks from the supply chain.

This work contributes to the security domain in cyber-physical manufacturing systems, with the focus on intrusion detection. The dataset collected during the experiments has been shared with the research community. The alert correlation methodology also contributes to cyber-physical systems, such as smart grid and connected vehicles, which requires enhanced security protection in today’s connected world.


Open Access

Included in

Engineering Commons