Quickest Change Detection Research Articles

Change detection in a dynamic stream of attributed networks

While anomaly detection in static networks has been extensively studied, only recently have researchers focused on dynamic networks. This trend is mainly due to the capacity of dynamic networks to represent complex physical, biological, cyber, and social systems. This article proposes a new methodology for modeling and monitoring dynamic attributed networks for quick detection of temporal changes in network structures. In this methodology, the generalized linear model (GLM) is used to model static attributed networks. This model is then combined with a state transition equation to capture the dynamic behavior of the system. Extended Kalman filter (EKF) is used as an online, recursive inference procedure to predict and update network parameters over time. In order to detect changes in the underlying mechanism of edge formation, prediction residuals are monitored through an exponentially weighted moving average (EWMA) control chart. The proposed modeling and monitoring procedure is examined through simulations for attributed binary and weighted networks. Email communication data from the Enron corporation is used as a case study to show how the method can be applied in real-world problems.

A spatial-adaptive sampling procedure for online monitoring of big data streams

With the improvement of data-acquisition technology, big data streams that involve continuous observations with high dimensionality and large volume frequently appear in modern applications, which poses significant challenges for statistical process control. In this article we consider the problem of online monitoring a class of big data streams where each data stream is associated with a spatial location. Our goal is to quickly detect shifts occurring in such big data streams when only partial information can be observed at each time and the out-of-control variables are clustered in a small and unknown region. To achieve this goal, we propose a novel spatial-adaptive sampling and monitoring (SASAM) procedure that aims to leverage the spatial information of the data streams for quick change detection. Specifically, the proposed sampling strategy will adaptively and intelligently integrate two seemingly contradictory ideas: (1) random sampling that quickly searches for possible out-of-control variables; and (2) directional sampling that focuses on highly suspicious out-of-control variables that may cluster in a small region. Simulation and real case studies show that the proposed method significantly outperforms the existing sampling strategy without taking the spatial information of the data streams into consideration.

Tracking Changes in Resilience and Level of Coordination in Terrorist Networks

The focus of this paper is on the activity profiles of certain networks, such as terrorist networks, that show frequent spurts and downfalls. Of particular interest is the quick detection of changes in two specific activity patterns corresponding to resilience and level of coordination in the network. Understanding changes in resilience and coordination could provide insights into the underlying organizational dynamics and aid in more informed decision-making. Prior work in tackling this problem is based on parametric approaches and relies on models developed with time-series analysis techniques, self-exciting hurdle models, or hidden Markov models. While these approaches detect spurts and downfalls reasonably accurately, they are all based on model learning—a task that is computationally difficult in practice because of the “rare” nature of terrorist attacks from a model learning perspective. In this paper, we pursue an alternate statistical nonparametric approach for spurt detection in activity profiles. Our approach is based on binning the count data of activity to form observation vectors that can be compared with each other. Motivated by a majorization theory framework, these vectors are then transformed via certain functionals and used in spurt detection and classification. While the parametric approaches often result in either a large number of missed detections of real changes or false alarms, the proposed approach is shown to result in a small number of missed detections and false alarms. Furthermore, since spurt detection is a problem of importance across multiple applications, the nonparametric nature of the approach makes it attractive in these applications.

Quickest Detection for Changes in Maximal kNN Coherence of Random Matrices

This paper addresses the problem of quickest detection of a change in the maximal coherence between columns of an $n\times p$ random matrix based on a sequence of matrix observations having a single unknown change point. The random matrix is assumed to have identically distributed rows and the maximal coherence is defined as the largest of the $p \atopwithdelims ()2$ correlation coefficients associated with any row. Likewise the $k$ nearest neighbor (kNN) coherence is defined as the $k$ -th largest of these correlation coefficients. The forms of the pre- and postchange distributions of the observed matrices are assumed to belong to the family of elliptically contoured densities with sparse dispersion matrices but are otherwise unknown. A nonparametric stopping rule is proposed that is based on the maximal kNN sample coherence between columns of each observed random matrix. This is a summary statistic that is related to a test of existence of a hub vertex in a sample correlation graph having degree at least $k$ . Performance bounds on the delay and false alarm performance of the proposed stopping rule are obtained in the purely high-dimensional regime where $p\rightarrow \infty$ and $n$ is fixed. When the prechange dispersion matrix is diagonal it is shown that, among all functions of the proposed summary statistic, the proposed stopping rule is asymptotically optimal under a minimax quickest change detection model as the stopping threshold approaches infinity. The theory developed also applies to sequential hypothesis testing and fixed sample size tests.

Evaluating the performance of the staircase and quick Change Detection methods in measuring perceptual learning

Evaluating the performance of the staircase and quick Change Detection methods in measuring perceptual learning

Assessing the trial-by-trial time course of perceptual sensitivity change in perceptual learning using the quick Change Detection method

Assessing the trial-by-trial time course of perceptual sensitivity change in perceptual learning using the quick Change Detection method

Accuracy and Reliability of Multiphase CTA Perfusion for Identifying Ischemic Core.

Acute stroke treatment requires simple, quick and accurate detection of early ischemic changes to facilitate treatment decisions guided by published selection criteria. The aim of this study was to determine the accuracy and reliability of multiphase computed tomography angiography (mCTA) perfusion hypoattenuation for detecting early severe ischemia. Non-contrast CT (NCCT), mCTA for regional leptomeningeal score (mCTA-rLMC), and mCTA perfusion lesion visibility (mCTA-arterial and mCTA-venous) were assessed blinded to clinical information in patients treated with endovascular therapy (EVT). The extent of early ischemia defined by regions of hypoattenuation was evaluated by the Alberta Stroke Program Early CT Score (ASPECTS). The ASPECTS scores were dichotomized based on the American Heart Association (AHA) guidelines for EVT selection, ASPECTS ≥6 vs. <6. The diagnostic accuracy was calculated by comparison to 24-h magnetic resonance imaging (MRI) or CT ASPECTS. Inter-observer reliability of NCCT and mCTA ASPECTS was evaluated. Machine learning models were used to predict the clinical follow-up outcome, e.g. National Institutes of Health Stroke scale (NIHSS) and modified Rankin scale (mRS) from baseline imaging data and patient information. A total of 89acute stroke patients (68 ± 15years of age) were analyzed (33TICI-0, 56TICI-2b or 3). Median baseline NIHSS was 17. The mCTA-venous had alarge effect on accurately identifying early ischemia when dichotomized for ASPECTS ≥6 vs <6 (likelihood ratio [LR+] > 10 vs. [LR-] < 0.29) compared to the moderate effect of NCCT ([LR+] = 6.7; [LR-] = 0.56) and mCTA-rLMC [(LR+ = 8.0; (LR-) = 0.83)]. The inter-observer reliability in mCTA-venous was almost perfect for all ASPECTS regions except the internal capsule. The machine learning support factor regression model identified mCTA-venous as the most important imaging covariate for predicting 24-h NIHSS and 90-day mRS. The assessment of mCTA-venous permits amore accurate detection of early ischemia than NCCT and mCTA-rLMC score and is predictive of clinical outcome. We would recommend the inclusion of mCTA perfusion lesion in future endovascular trials aiming at extending current AHA guidelines for EVT in stroke patients with low ASPECTS.

Asymptotic statistical properties of communication-efficient quickest detection schemes in sensor networks

The quickest change detection problem is studied in a general context of monitoring a large number K of data streams in sensor networks when the “trigger event” may affect different sensors differently. In particular, the occurring event might affect some unknown, but not necessarily all, sensors and also could have an immediate or delayed impact on those affected sensors. Motivated by censoring sensor networks, we develop scalable communication-efficient schemes based on the sum of those local cumulative sum (CUSUM) statistics that are “large” under either hard, soft, or order thresholding rules. Moreover, we provide the detection delay analysis of these communication-efficient schemes in the context of monitoring K independent data streams and establish their asymptotic statistical properties under two regimes: one is the classical asymptotic regime when the dimension K is fixed, and the other is the modern asymptotic regime when the dimension K goes to Our theoretical results illustrate the deep connections between communication efficiency and statistical efficiency.

Efficient and Low Latency Detection of Intruders in Mobile Active Authentication

Active authentication (AA) refers to the problem of continuously verifying the identity of a mobile device user for the purpose of securing the device. We address the problem of quickly detecting intrusions with lower false detection rates in mobile AA systems with higher resource efficiency. Bayesian and MiniMax versions of the quickest change detection (QCD) algorithms are introduced to quickly detect intrusions in mobile AA systems. These algorithms are extended with an update rule to facilitate low-frequency sensing which leads to low utilization of resources. Effectiveness of the proposed framework is demonstrated using three publicly available unconstrained face and touch gesture-based AA datasets. It is shown that the proposed QCD-based intrusion detection methods can perform better than many state-of-the-art AA methods in terms of latency and low false detection rates. Furthermore, it is shown that employing the proposed resource-efficient extension further improves the performance of the QCD-based setup.

UAV-BASED 3D URBAN ENVIRONMENT MONITORING

Abstract. Unmanned Aerial Vehicle (UAV) based remote sensing can be used to make three-dimensions (3D) mapping with great flexibility, besides the ability to provide high resolution images. In this paper we propose a quick-change detection method on UAV images by combining altitude from Digital Surface Model (DSM) and texture analysis from images. Cases of UAV images with and without georeferencing are both considered. Research results show that the accuracy of change detection can be enhanced with georeferencing procedure, and the accuracy and precision of change detection on UAV images which are collected both vertically and obliquely but without georeferencing also have a good performance.

Quickest Change Detection in Adaptive Censoring Sensor Networks

The problem of quickest change detection with communication rate constraints is studied. A network of wireless sensors with limited computation capability monitors the environment and sends observations to a fusion center via wireless channels. At an unknown time instant, the distributions of observations at all the sensor nodes change simultaneously. Due to limited energy, the sensors cannot transmit at all the time instants. The objective is to detect the change at the fusion center as quickly as possible, subject to constraints on false detection and average communication rate between the sensors and the fusion center. A minimax formulation is proposed. The cumulative sum (CuSum) algorithm is used at the fusion center and censoring strategies are used at the sensor nodes. The censoring strategies, which are adaptive to the CuSum statistic, are fed back by the fusion center. The sensors only send observations that fall into prescribed sets to the fusion center. This CuSum adaptive censoring (CuSum-AC) algorithm is proved to be an equalizer rule and to be globally asymptotically optimal for any positive communication rate constraint, as the average run length to false alarm goes to infinity. It is also shown, by numerical examples, that the CuSum-AC algorithm provides a suitable trade-off between the detection performance and the communication rate.

Exploration of Singular Spectrum Analysis for Online Anomaly Detection in CRNs

Cognitive radio networks (CRNs) have been recognized as a promising technology that allows secondary users (SUs) extensively explore spectrum resource usage efficiency, while not introducing interference to licensed users. Due to the unregulated wireless network environment, CRNs are susceptible to various malicious entities. Thus, it is critical to detect anomalies in the first place. However, from the perspective of intrinsic features of CRNs, there is hardly in existence of an universal applicable anomaly detection scheme. Singular Spectrum Analysis (SSA) has been theoretically proven an optimal approach for accurate and quick detection of changes in the characteristics of a running (random) process. In addition, SSA is a model-free method and no parametric models have to be assumed for different types of anomalies, which makes it a universal anomaly detection scheme. In this paper, we introduce an adaptive parameter and component selection mechanism based on coherence for basic SSA method, upon which we built up a sliding window online anomaly detector in CRNs. Our experimental results indicate great accuracy of the SSA-based anomaly detector for multiple anomalies.

Misspecified and Asymptotically Minimax Robust Quickest Change Detection

We investigate the quickest detection of an unknown change in the distribution of a stochastic process generating independent and identically distributed observations. We develop new bounds on the performance of misspecified cumulative sum (CUSUM) rules, and pose minimax robust versions of the popular Lorden and Pollak criteria with polynomial (or higher order moment) detection delay penalties. By exploiting our results for misspecified CUSUM rules, we identify solutions to our robust quickest change detection problems in the asymptotic regime of few false alarms. In contrast to previous robust quickest change detection treatments, our asymptotic results hold under relaxed conditions on the uncertainty sets of possible prechange and postchange distributions. We illustrate our results in simulations and apply them to the problem of detecting target manoeuvres in low signal-to-noise ratio settings (i.e., dim-target manoeuvre detection).

Minimax Robust Quickest Change Detection With Exponential Delay Penalties

Quickly detecting changes in the statistical behavior of measurements is important in many applications of control engineering involving fault detection and process monitoring. In this letter, we pose and solve minimax robust Lorden and Bayesian quickest change detection problems for situations where the cost of detection delays compounds exponentially. We show that the detection rules that solve our robust quickest change detection problems are also the rules that solve the standard (non-robust) problems specified by least favorable distributions from uncertainty classes of possible distributions that satisfy a stochastic boundedness condition. In contrast to previous robust quickest change detection results with nonlinear detection delay penalties, our results with exponential delay penalties are exact (i.e., they hold for any false alarm constraint and not only in the asymptotic regime of few false alarms). We illustrate our results through simulations.

Minimax optimality of Shiryaev-Roberts procedure for quickest drift change detection of a Brownian motion

ABSTRACTThe problem of detecting a change in the drift of a Brownian motion is considered. The change point is assumed to have a modified exponential prior distribution with unknown parameters. A worst-case analysis with respect to these parameters is adopted leading to a min–max problem formulation. Analytical and numerical justifications are provided toward establishing that the Shiryaev-Roberts procedure with a specially designed starting point is exactly optimal for the proposed mathematical setup.

Statistical Power System Line Outage Detection Under Transient Dynamics

A quickest change detection (QCD) algorithm is proposed to address the problem of detecting and identifying line outages in a power system. The statistics of electricity demand are assumed to be known and propagated through a linearized model of the equations describing the power flow balance at each node of the network. The proposed algorithm incorporates the transient dynamics of the power system following a line outage and is applied to the measurements of voltage phase angles, which are collected using phasor measurement units. This adaptive algorithm is developed by treating the QCD problem as a dynamic composite hypothesis testing problem. It is shown to have superior performance compared with other line outage detection algorithms previously proposed in the literature. Case studies demonstrating this gain in performance are illustrated through the IEEE 118-bus test system.

Quickest detection in the Wiener disorder problem with post-change uncertainty

We consider the problem of quickest detection of an abrupt change when there is uncertainty about the post-change distribution. In particular, we examine this problem in the continuous-time Wiener model where the drift of observations changes from zero to a random drift with a prescribed discrete distribution. We set up the problem as a stochastic optimization in which the objective is to minimize a measure of detection delay subject to a constraint on frequency of false alarms. We design a novel composite stopping rule and prove that it is asymptotically optimal of third order under a weighted Lorden’s criterion for detection delay. We also develop the strategy to identify the post-change drift and analyze the conditional identification error asymptotically. Our composite rules are based on CUSUM stopping times, as well as their reaction periods, namely the times between the last reset of the CUSUM statistic process and the CUSUM alarm. The established results shed new light on the performance of CUSUM strategies under model uncertainty and offer strong asymptotic optimality results in this framework.

Quickest Change Detection With Observation Scheduling

The quickest change detection problem is to detect an abrupt change event as quickly as possible subject to constraints on false detection. Unlike the classical problem, where the decision maker can access only one sequence of observations, in this paper, the decision maker chooses one of two different sequences of observations at each time instant. The information quality and sampling cost of the two sequences of observations are different. We present an asymptotically optimal joint design of observation scheduling policy and stopping time such that the detection delay is minimized subject to constraints on both average run length to false alarm (ARLFA) and average cost per sample. The observation scheduling policy has a threshold structure and the detection scheme is a variant of the cumulative sum test where the detection statistic stochastically crosses the threshold that is used to switch observation modes. We further study the decentralized case in a multi-channel setting. We show that if each sensor uses the proposed observation scheduling policy locally and the fusion center uses the N <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">sum</sub> algorithm, by which the center declares the change when the sum of the sensors' local detection statistics crosses a certain threshold, the detection delay is asymptotically minimized for any possible combination of the affected sensors subject to constraints on both global ARLFA and average cost per sample at each sensor node. Numerical examples are given to illustrate the main results.

Detecting a Suddenly Arriving Dynamic Profile of Finite Duration

This paper addresses the detection of a suddenly arriving dynamic profile of a finite duration often called a transient change. In contrast to the traditional abrupt change detection, where the post-change period is assumed to be infinitely long, the detection of a suddenly arriving transient change should be done before it disappears. The detection of transient changes after their disappearance is considered as missed. Hence, the traditional quickest change detection criterion, minimizing the average detection delays provided a prescribed false alarm rate, is compromised. The proposed optimality criterion minimizes the worst case probability of missed detection provided that the worst case probability of false alarm during a certain period is upper bounded. A suboptimal CUSUM-type transient change detection algorithm, based on a subclass of truncated Sequential Probability Ratio Tests, is proposed. The optimization of the proposed algorithm in this subclass leads to a specially designed Finite Moving Average Test. The proposed method is analyzed theoretically and by simulation. A special attention is paid to the case of Gaussian observations with a dynamic profile.

Bayesian Quickest Change Detection for Active Sensors

We consider the energy efficient quickest change detection problem for active sensors, which use radar, ultrasound, or optical sensing to control the information content in the sequentially collected noisy data. Controlling the extent of information contained in the collected data would lead to a cost or energy expenditure, where the cost would naturally increase as the extent of contained information increases. For such a sensor, our objective is to design control policies that optimally trade off performance metrics, such as detection delay, average cost expended in control, and probability of false alarm. We consider this problem in the setting of Bayesian quickest change detection. We propose two simple heuristic threshold policies, for which the total cost is shown to be close to the optimal through simulations and numerical studies.