Stavros Ntalampiras

A Cognitive Fault Diagnosis System for Distributed Sensor Networks

This paper introduces a novel cognitive fault diagnosis system (FDS) for distributed sensor networks that takes advantage of spatial and temporal relationships among sensors. The proposed FDS relies on a suitable functional graph representation of the network and a two-layer hierarchical architecture designed to promptly detect and isolate faults. The lower processing layer exploits a novel change detection test (CDT) based on hidden Markov models (HMMs) configured to detect variations in the relationships between couples of sensors. HMMs work in the parameter space of linear time-invariant dynamic systems, approximating, over time, the relationship between two sensors; changes in the approximating model are detected by inspecting the HMM likelihood. Information provided by the CDT layer is then passed to the cognitive one, which, by exploiting the graph representation of the network, aggregates information to discriminate among faults, changes in the environment, and false positives induced by the model bias of the HMMs.

Universal background modeling for acoustic surveillance of urban traffic

Abstract Traffic congestion in modern cities is an increasing problem having significant consequences in our daily lives. This work proposes a non-intrusive, passive monitoring framework based on the acoustic modality which can be used either autonomously or as a part of a multimodal system and provide valuable information to an intelligent transportation system. We consider a large number of audio classes which are typically encountered in urban areas. We introduce a combination of a powerful audio representation mechanism based on time, frequency and wavelet domain features with universal background modeling which leads to higher recognition accuracies and detection rates (in terms of false alarm and miss probability rates) with respect to commonly employed methodologies. The basic advantage of a class-specific model derived using the universal background modeling logic is its tolerance to data which belong to other sound classes. Another important feature of the proposed system is its ability to detect crash incidents, which apart from their catastrophic impact on human life and property, have negative consequences on the traffic flow. Our experiments are based on the concurrent usage of professional sound effect collections which include audio recordings of high quality. We thoroughly examine the performance of the proposed system on isolated sound events as well as continuous audio streams using confusion matrices and detection error trade-off curves.

An HMM-based change detection method for intelligent embedded sensors

In this work we address the problem of automatically detecting changes either induced by faults or concept drifts in data streams coming from multi-sensor units. The proposed methodology is based on the fact that the relationships among different sensor measurements follow a probabilistic pattern sequence when normal data, i.e. data which do not present a change, are observed. Differently, when a change in the process generating the data occurs the probabilistic pattern sequence is modified. The relationship between two generic data streams is modelled through a sequence of linear dynamic time-invariant models whose trained coefficients are used as features feeding a Hidden Markov Model (HMM) which, in turn, extracts the pattern structure. Change detection is achieved by thresholding the log-likelihood value associated with incoming new patterns, hence comparing the affinity between the structure of new acquisitions with that learned through the HMM. Experiments on both artificial and real data demonstrate the appreciable performance of the method both in terms of detection delay, false positive and false negative rates.

Fault Identification in Distributed Sensor Networks Based on Universal Probabilistic Modeling

This paper proposes a holistic modeling scheme for fault identification in distributed sensor networks. The proposed scheme is based on modeling the relationship between two datastreams by means of a hiddenMarkov model (HMM) trained on the parameters of linear time-invariant dynamic systems, which estimate the specific relationship over consecutive time windows. Every system state, including the nominal one, is represented by an HMM and the novel data are categorized according to the model producing the highest likelihood. The system is able to understand whether the novel data belong to the fault dictionary, are fault-free, or represent a new fault type. We extensively evaluated the discrimination capabilities of the proposed approach and contrasted it with a multilayer perceptron using data coming from the Barcelona water distribution network. Nine system states are present in the dataset and the recognition rates are provided in the confusion matrix form.

Automatic identification of integrity attacks in cyber-physical systems

The cyber-layer existing in modern infrastructures opened the door to the emerging threat of cyber-attacks.This paper proposes a novel methodology for automatic identification of integrity attacks.The identification is performed by capturing the unique characteristics of each attack in the spectral and wavelet domains.The pattern of each cyber attack is captured by recognition algorithms of different modelling properties.The efficacy of the proposed approach is demonstrated in the Smart Grid domain. Modern society relies on the availability and smooth operation of complex engineering systems, such as electric power systems, water distributions networks, etc. which due to the recent advancements in information and communication technologies (ICT) are usually controlled by means of a cyber-layer. This design may potentially improve the usage of the components of the cyber-physical system (CPS), however further protection is needed due to the emerging threat of cyber-attacks. These may degrade the quality of the communicated information which is of fundamental importance in the decision making process.This paper proposes a novel methodology for automatic identification of the type of the integrity attack affecting a CPS. We designed a feature set for capturing the characteristics of each attack in the spectral and wavelet domains while its distribution is learned by pattern recognition algorithms of different modelling properties customized for the specific application scenario. In addition a novelty detection component is incorporated for dealing with previously unseen types of attacks. The proposed approach is applied onto data coming from the IEEE-9 bus model achieving promising identification performance.

Model-Free Fault Detection and Isolation in Large-Scale Cyber-Physical Systems

Detecting and isolating faults in cyber-physical systems (CPSs), e.g., critical infrastructures, smart buildings/cities, and the internet-of-things, are tasks that do generally scale badly with the CPS size. This work introduces a model-free fault detection and diagnosis system (FDDS) designed, having in mind scalability issues, so as to be able to detect and isolate faults in CPSs characterised by a large number of sensors. Following the model-free approach, the proposed FDDS learns the nominal fault-free conditions of the large-scale CPS autonomously by exploiting the temporal and spatial relationships existing among sensor data. The novelties in this paper reside in 1) a clustering method proposed to partition the large-scale CPS into groups of highly correlated sensors in order to grant scalability of the proposed FDDS, and 2) the design of model- and fault-free mechanisms to detect and isolate multiple sensor faults, and disambiguate between sensor faults and time variance of the physical phenomenon the cyber layer of CPS inspects.

PROMETHEUS: heterogeneous sensor database in support of research on human behavioral patterns in unrestricted environments

The multi-modal multi-sensor PROMETHEUS database was created in support of research and development activities [PROMETHEUS (FP7-ICT-214901): http://www.prometheus-FP7.eu] aiming at the creation of a framework for monitoring and interpretation of human behaviors in unrestricted indoor and outdoor environments. The distinctiveness of the PROMETHEUS database comes from the unique sensor sets, used in the various recording scenarios, but also from the database design, which covers a range of real-world applications, correlated to smart-home automation and indoors/outdoors surveillance of public areas. Numerous single-person and multi-person scenarios, but also scenarios with interactions between groups of people, motivated by these applications were implemented with the help of skilled actors and supernumerary personnel. In these scenarios, the actors and personnel were instructed to implement a range of typical and atypical behaviors, and simulations of emergency and crisis situations. In summary, the database contains more than 4 h of synchronized recordings from heterogeneous sensors (an infrared motion detection sensor, thermal imaging cameras, overview/surveillance video cameras, close-view video cameras, a 3D camera, a stereoscopic camera, a general-purpose camcoder, microphone arrays, and motion capture equipment) collected in common setups, simulating smart-home environment, airport, and ATM security environment. Selected scenes of the database were annotated for the needs of human detection and tracking. The entire audio part of the database was annotated for the needs of sound event detection, sound source enumeration, emotion recognition, etc.

Temporal/spatial model-based fault diagnosis vs. Hidden Markov models change detection method: Application to the Barcelona water network

This paper deals with a comparison of two different fault diagnosis frameworks. The first method is based on a temporal/spatial model-based analysis by exploiting a-priori information about the system under study, so fault detection is based on monitoring the residuals of combined spatial and time series models obtained from the network. The second method aims at characterizing and detecting changes in the probabilistic pattern sequence of data coming from the network. Relationships between data streams are modelled through sequences of linear dynamic time-invariant models whose trained coefficients are used to feed a Hidden Markov Model (HMM). When the pattern structure of incoming data cannot be explained by the trained HMM, a change is detected. Here, the performance obtained from this two distinct approaches is examined by using a dataset coming from the Barcelona water transport network.

Model ensemble for an effective on-line reconstruction of missing data in sensor networks

The literature has shown that model ensemble techniques are particularly effective to solve regression/classification applications by providing, given a suitable aggregation mechanism, a better generalization ability than the generic model of the ensemble. However, only few recent results consider the use of ensembles for a time-dependent framework, with focus on time-series forecasting. Here, we propose the use of ensemble of models to an on-line reconstruction of missing data coming from a sensor network. Reconstructing missing data is of paramount importance for any further data processing and must be carried out on-line not to introduce unnecessary latency when data lead to a decision or control action. The ensemble is designed by both exploiting temporal and spatial dependencies existing among the sensors composing the network. An effective aggregation mechanism is proposed for the considered models to improve the generalization ability of the ensemble. Results demonstrate the effectiveness of the proposed approach in reconstructing missing data.

Fault Diagnosis for Smart Grids in Pragmatic Conditions

Due to the advancements of electrical networks, the operators are able to employ a gamut of information for assessing the state of the infrastructure facilitating diagnosis of potential malfunctions appearing in one or more components of the grid. This paper presents a cognitive fault diagnosis framework for smart grids (SGs) which exploits the temporal and functional relationships existing within the datastreams coming from the nodes of the network. The protection of SGs can rely not only on conventional techniques (e.g., circuit breakers) but also on processing information which is available, thanks to the information and communication layer. We propose a framework which is able to autonomously learn the model of the nominal state using the respective data by means of hidden Markov models operating in the parameter space of linear time-invariant models. Subsequently, the framework is able to detect data not belonging to the nominal state and localize the potential fault at the cognitive level. The isolation is based on a graph representation of the SG revealing the correlations among the nodes based on the Granger causality. We conducted thorough experiments on the IEEE-9 bus system model achieving encouraging results in terms of false positive/negative rate and detection/isolation delay.