Machine Learning Methods for Attack Detection in the Smart Grid

Özay, Mete; Esnaola, Iñaki; Vural, Fatoş T. Yarman; Kulkarni, Sanjeev R.; Poor, H. Vincent

doi:10.1109/tnnls.2015.2404803

Cited by 532 publications

(255 citation statements)

References 39 publications

Supporting

Mentioning

254

Contrasting

Order By: Relevance

“…The family of machine learning algorithms can be categorized based on their functionality and structure [1], yielding regression algorithms, decision tree algorithms, Bayesian algorithms, clustering algorithms, and artificial neural networks, just to name a few. In this article, we clarify the machine learning techniques from two perspectives: parametric/non-parametric learning and supervised/unsupervised/reinforcement learning, as illustrated in Non-parametric learning methods: In contrast to parametric learning methods, the nonparametric learning methods [1], [2], [13] are not specified a priori, but are determined from the available data. Examples include kernel estimator, k-nearest neighbors, and decision trees, just to name a few.…”

Section: Machine Learning Paradigms For Intelligent Authenticationmentioning

confidence: 99%

Machine Learning for Intelligent Authentication in 5G and Beyond Wireless Networks

Fang

Wang

Tomasin

2019

IEEE Wireless Commun.

123

View full text Add to dashboard Cite

The fifth generation (5G) and beyond wireless networks are critical to support diverse vertical applications by connecting heterogeneous devices and machines, which directly increase vulnerability for various spoofing attacks. Conventional cryptographic and physical layer authentication techniques are facing some challenges in complex dynamic wireless environments, including significant security overhead, low reliability, as well as difficulties in pre-designing a precise authentication model, providing continuous protection, and learning time-varying attributes. In this article, we envision new authentication approaches based on machine learning techniques by opportunistically leveraging physical layer attributes, and introduce intelligence to authentication for more efficient security provisioning. Machine learning paradigms for intelligent authentication design are presented, namely for parametric/non-parametric and supervised/unsupervised/reinforcement learning algorithms. In a nutshell, the machine learning-based intelligent authentication approaches utilize specific features in the multi-dimensional domain for achieving cost-effective, more reliable, model-free, continuous, and situation-aware device validation under unknown network conditions and unpredictable dynamics. 2 heterogeneous devices and machines [1]. The dramatically growing number of low-cost devices and access points with increased mobility and heterogeneity leads to the extremely complex and dynamic environment of 5G-and-beyond wireless networks. Specifically, due to the open broadcast nature of radio signal propagation, widely adopted standardized transmission protocols, and intermittent communication characteristic, wireless communications are extremely vulnerable to interception and spoofing attacks [2]. As shown in the conventional 'Alice-Bob-Eve' model of Fig. 1, Alice and Bob are legitimate devices and communicate with each other in the presence of a Spoofer (Eve). This adversary dynamically intends to intercept legitimate communications and to impersonate Alice for obtaining illegal advantages from Bob. The main objective of Bob is to uniquely and unambiguously identify Alice by authentication techniques. Multi-dimensional information received at Bob Time Spoofing Signal Legitimate Signal Scattering cluster Scattering cluster Scattering cluster Eve (Spoofer) Alice Bob Process-related information Physical information Fig. 1: Alice-Bob-Eve model in the complex time-varying environment. Bob identifies Alice from Spoofer based on the multi-dimensional received information, which may be related to communication links, physical environment, and communication process, just to name a few.Although cryptographic techniques have been widely studied for authentication, they may fall short of the desired performance in many emerging scenarios of 5G-and-beyond wireless networks [2]. The fundamental weaknesses of conventional cryptography techniques are the increasing latencies, communication and computation overheads for achieving better security pe...

show abstract

Section: Machine Learning Paradigms For Intelligent Authenticationmentioning

confidence: 99%

Machine Learning for Intelligent Authentication in 5G and Beyond Wireless Networks

Fang

Wang

Tomasin

2019

IEEE Wireless Commun.

123

View full text Add to dashboard Cite

show abstract

“…Smart grid compromised device detection: In general, the topic of compromised devices has not been extensively studied in the literature. In most cases, researchers focus on proposing anomaly detection mechanisms [51] for di erent types of a acks in the smart grid [24,30,66,78], without particularizing on the a ack sources (e.g., compromised devices). In a few cases, however, the behavior of the smart grid device is considered.…”

Section: Related Workmentioning

confidence: 99%

A System-level Behavioral Detection Framework for Compromised CPS Devices

Babun

Aksu

Uluagac

2019

ACM Trans. Cyber-Phys. Syst.

View full text Add to dashboard Cite

Cyber-Physical Systems (CPS) play a signi cant role in our critical infrastructure networks from power-distribution to utility networks. e emerging smart-grid concept is a compelling critical CPS infrastructure that relies on two-way communications between smart devices to increase e ciency, enhance reliability, and reduce costs. However, compromised devices in the smart grid poses several security challenges. Consequences of propagating fake data or stealing sensitive smart grid information via compromised devices are costly. Hence, early behavioral detection of compromised devices is critical for protecting the smart grid's components and data. To address these concerns, in this paper, we introduce a novel and con gurable system-level framework to identify compromised smart grid devices. e framework combines system and function call tracing techniques with signal processing and statistical analysis to detect compromised devices based on their behavioral characteristics. We measure the e cacy of our framework with a realistic smart grid substation testbed that includes both resource-limited and resource-rich devices. In total, using our framework, we analyze six di erent types of compromised device scenarios with di erent resources and a ack payloads. To the best of our knowledge, the proposed framework is the rst in detecting compromised CPS smart grid devices with system and function-level call tracing techniques. e experimental results reveal an excellent rate for the detection of compromised devices. Speci cally, performance metrics include accuracy values between 95% and 99% for the di erent a ack scenarios. Finally, the performance analysis demonstrates that the use of the proposed framework has minimal overhead on the smart grid devices' computing resources.

show abstract

“…Within the given configuration, the system allows us to define the range of knowledge for learning, choose the learning algorithm, and set all necessary parameters. Before adding the learned rules to the knowledge base, the user may review and verify the new knowledge [22].…”

Section: Machine Learningmentioning

confidence: 99%

Reasoning Algorithm for the Creative Decision Support System Integrating Inference and Machine Learning

Wilk-Kołodziejczyk

2017

csci

View full text Add to dashboard Cite

In this paper, a reasoning algorithm for a creative decision support system is proposed. It allows for the integration of inference and machine learning algorithms. The execution of the learning algorithm is automatic since it is formalized by applying a complex inference rule, which generates intrinsically new knowledge using the facts already stored in the knowledge base as training data. This new knowledge may be used in the same inference chain to derive a decision. A solution of this type makes the reasoning process more creative and also allows for its continuation in cases when the knowledge base does not have the appropriate explicitly encoded knowledge. In this paper, an appropriate knowledge representation and inference model are proposed. Experimental verication is performed on the decision support system operating in a casting domain.

show abstract

Machine Learning Methods for Attack Detection in the Smart Grid

Cited by 532 publications

References 39 publications

Machine Learning for Intelligent Authentication in 5G and Beyond Wireless Networks

Machine Learning for Intelligent Authentication in 5G and Beyond Wireless Networks

A System-level Behavioral Detection Framework for Compromised CPS Devices

Reasoning Algorithm for the Creative Decision Support System Integrating Inference and Machine Learning

Contact Info

Product

Resources

About