Detecting anomalies and attacks in smart cyber-physical systems are of paramount importance owing to their growing prominence in controlling critical systems. However, this is a challenging task due to the heterogeneity and variety of components of a CPS, and the complex relationships between sensed values and potential attacks or anomalies. Such complex relationships are results of physical constraints and domain norms which exist in many CPS domains. In this article, we proposeABATe, anAutomaticBehavioralAbstractionTechnique based on neural networks for detecting anomalies in smart cyber-physical systems. Unlike traditional techniques which abstract the statistical properties of different sensor values,ABATelearns complex relationships between event vectors from normal operational data available in abundance with smart CPS and uses this abstracted model to detect anomalies.ABATedetected more than 88 percent of attacks in the publicly available SWaT dataset featuring data from a scaled down sewage water treatment plant with a very low false positive rate of 1 percent. We also evaluated our technique's ability to capture domain semantics and multi-domain adaptability using a real-world automotive dataset, as well as a synthetic dataset.

Machine learning algorithms used to detect attacks are limited by the fact that they cannot incorporate the back-ground knowledge that an analyst has. This limits their suitability in detecting new attacks. Reinforcement learning is different from traditional machine learning algorithms used in the cybersecurity domain. Compared to traditional ML algorithms, reinforcement learning does not need a mapping of the input-output space or a specific user-defined metric to compare data points. This is important for the cybersecurity domain, especially for malware detection and mitigation, as not all problems have a single, known, correct answer. Often, security researchers have to resort to guided trial and error to understand the presence of a malware and mitigate it.In this paper, we incorporate prior knowledge, represented as Cybersecurity Knowledge Graphs (CKGs), to guide the exploration of an RL algorithm to detect malware. CKGs capture semantic relationships between cyber-entities, including that mined from open source. Instead of trying out random guesses and observing the change in the environment, we aim to take the help of verified knowledge about cyber-attack to guide our reinforcement learning algorithm to effectively identify ways to detect the presence of malicious filenames so that they can be deleted to mitigate a cyber-attack. We show that such a guided system outperforms a base RL system in detecting malware.

To effectively identify and filter out attacks from known sources like botnets, spammers, virus infected systems etc., organizations increasingly procure services that determine the reputation of IP addresses. Adoption of encryption techniques like TLS 1.2 and 1.3 aggravate this cause, owing to the higher cost of decryption needed for examining traffic contents. Currently, most IP reputation services provide blacklists by analyzing malware and spam records. However, newer but similar IP addresses used by the same attackers need not be present in such lists and attacks from them will get bypassed. In this paper, we present Dynamic Attribute based Reputation (DAbR), a euclidean distance based technique, to generate reputation scores for IP addresses by assimilating meta-data from known bad IP addresses. This approach is based on our observation that many bad IP's share similar attributes and the requirement for a lightweight technique for reputation scoring. DAbR generates reputation scores for IP addresses on a 0-10 scale which represents its trustworthiness based on known bad IP address attributes. The reputation scores when used in conjunction with a policy enforcement module, can provide high performance and non privacy-invasive malicious traffic filtering. To evaluate DAbR, we calculated reputation scores on a dataset of 87k IP addresses and used them to classify IP addresses as good/bad based on a threshold. An F-1 score of 78% in this classification task demonstrates our technique's performance.

According to recent media reports, there has been a surge in the number of devices that are being connected to the Internet. The Internet of Things (IoT), also referred to as Cyber-Physical Systems, is a collection of physical entities with computational and communication capabilities. The storage and computing power of these devices is often limited and their designs currently focus on ensuring functionality and largely ignore other requirements, including security and privacy concerns. We present the design of a framework that allows IoT devices to capture, represent, reason with, and enforce information sharing policies. We use Semantic Web technologies to represent the policies, the information to be shared or protected, and the IoT device context. We discuss use-cases where our design will help in creating an "intelligent" IoT device and ensuring data security and privacy using context-sensitive information sharing policies.

Vehicles can be considered as a specialized form of Cyber Physical Systems with sensors, ECU''s and actuators working together to produce a coherent behavior. With the advent of external connectivity, a larger attack surface has opened up which not only affects the passengers inside vehicles, but also people around them. One of the main causes of this increased attack surface is because of the advanced systems built on top of old and less secure common bus frameworks which lacks basic authentication mechanisms. To make such systems more secure, we approach this issue as a data analytic problem that can detect anomalous states. To accomplish that we collected data flowing between different components from real vehicles and using a Hidden Markov Model, we detect malicious behaviors and issue alerts, while a vehicle is in operation. Our evaluations using single parameter and two parameters together provide enough evidence that such techniques could be successfully used to detect anomalies in vehicles. Moreover our method could be used in new vehicles as well as older ones.