Transportation networks are considered one of the critical physical infrastructures for resilient cities (cyber-physical systems). In efforts to minimize adverse effects that come with the advancement of vehicular technologies, various governmental agencies, such as the U.S. Department of Homeland Security and the National Highway Traffic Safety Administration (NHTSA), work together. This paper develops belief-network-based attack modeling at signalized traffic networks under connected vehicle and intelligent signals frameworks. For different types of cyber attacks, defined in the literature, risk areas and impacts of attacks are evaluated. Vulnerability scores, technically based on the selected metrics, are calculated for signal controllers. In addition, the effect of having redundant traffic sensing systems on intersection performance measures is demonstrated in terms of average queue length differences.Resilience of critical infrastructures is defined as their ability to withstand an upsetting event, deliver essential levels of service during it, and recover quickly after it. With the increase of connected systems, cyber attacks that can target critical infrastructure systems are becoming more troubling. Transportation networks are considered one of the critical physical infrastructures for resilient cities (cyber-physical systems [CPSs] [1]). According to recent reports (2-4), several benefits are foreseen from upcoming technologies such as connected and autonomous vehicles (CAVs), including up to 80% reduction in fatalities from multi-vehicle crashes and prevention of the majority of human-error-related incidents, which takes out about 94% of all incidents. These intelligent applications, however, come at a price; for example, in 2015 alone 1.5 million vehicles were recalled because of cybersecurity vulnerabilities. NHTSA's current research focuses on CAVs that are heavily involved in secure implementations which will enable the field and its technology experts to harness efficient, reliable, and secure system design (3). Some of these topics can be listed as anomaly-based intrusion detection systems, cybersecurity of firmware updates, cybersecurity on heavy vehicles, vehicle-to-vehicle (V2V) communication interfaces, and trusted vehicle-to-everything (V2X) communications (5). The main goals for any critical infrastructure are quick detection of attacks and rapid mitigation efforts (6). There are many attack types, of which some can be resolved via detection and some require redundant systems and sensors. In intelligent transportation systems (ITS), to increase security and resiliency in case of possible attacks or benign system errors during different events, research is likewise needed into detection using various sensors and data types. Research is also necessary to enhance confidence in sensor readings by checking consistency with other sensors and information sources as well as validating control system commands (7). This paper investigates attack modeling and impacts on intelligent signals. For differe...

Attack graphs used in network security analysis are analyzed to determine sequences of exploits that lead to successful acquisition of privileges or data at critical assets. An attack graph edge corresponds to a vulnerability, tacitly assuming a connection exists and tacitly assuming the vulnerability is known to exist. In this thesis, we explore use of uncertain graphs to extend the paradigm to include lack of certainty in connection and/or existence of a vulnerability. We extend the standard notion of uncertain graph (where the existence of each edge is probabilistically independent) however, as significant correlations on edge existence probabilities exist in practice, owing to common underlying causes for disconnectivity and/or presence of vulnerabilities. Our extension describes each edge probability as a Boolean expression of independent indicator random variables. This thesis (i) shows that this formalism is maximally descriptive in the sense that it can describe any joint probability distribution function of edge existence, (ii) shows that when these Boolean expressions are monotone then we can easily perform uncertainty analysis of edge probabilities, and (iii) uses these results to model a partial attack graph of the Stuxnet worm and a small enterprise network and to answer important security-related questions in a probabilistic manner.

No abstract

In the IoT, massive distribution and long physical lifetimes will disrupt the "penetrate and patch" security paradigm that helps mitigate the consequences of the vulnerabilities endemic in individual systems. In this paper, we examine what will happen in the IoT if we build its systems the same way. We collect data and model the vulnerability blooms and patching delays in historical systems. We present the models and discuss future IoT networks where similar blooms happen but patching does not. We discuss initial results, and our plans to extend the models to look more deeply at these questions in our future work.

No abstract