Cyber threat intelligence and security informatics play critical roles in the identification of the principal influencers and threats associated with criminal activity and extremism in web-based communities. This paper presents an effective, dynamic learning framework utilizing domain specific context information to design a novel classification model that can robustly identify malicious social media posts (e.g. enrollment propaganda for extremist groups) with expressions of extremism or criminal intent. Research towards automated identification of extreme online posts and their associated key suspects and threats faces numerous challenges, 1) Online data, particularly social media data, originated from numerous independent and heterogeneous sources are largely unstructured. 2) The tactics, techniques, and procedures (TTPs) of criminal activity and extremism are constantly evolving. 3) There are limited ground truth data to support the development of effective classification technologies. In this paper, we present a human-machine collaborative, semi-supervised learning system that can efficiently and effectively identify malicious social media posts in presence of these challenges. Our system and framework develops an initial classifier from limited annotated data and in an interactive manner evolves dynamically into a sophisticated model using shortlisted relevant samples, identified via a graph-based optimization method, solvable by maximum flow algorithm. This same method also may be used to refine the classifier as TTPs evolve. Under this framework, the classifier performance converges faster using roughly 1-2 orders of magnitude fewer annotated samples, as compared to fully supervised solutions, resulting in a reasonably acceptable accuracy of nearly 80%. We validate our framework using a large collection of English and non-English flagged words extracted from three web-based forums and manually verified by multiple independent annotators.

The objective of a news veracity detection system is to identify various types of potentially misleading or false information, typically in a digital platform. A critical challenge in this scenario is that there are large volumes of data available online. However, obtaining samples with annotations (i.e. ground-truth labels) is difficult and a known limiting factor for many data analytic tasks including the current problem of news veracity detection. In this paper, we propose a human-machine collaborative learning system to evaluate the veracity of a news content, with a limited amount of annotated data samples. In a semi-supervised scenario, an initial classifier is learnt on a small, limited amount of the annotated data followed by an interactive approach to gradually update the model by shortlisting only relevant samples from the large pool of unlabeled data that are most likely to improve the classifier performance. Our prioritized active learning solution achieves faster convergence in terms of the classification performance, while requiring about 1–2 orders of magnitude fewer annotated samples compared to fully supervised solutions to attain a reasonably acceptable accuracy of nearly 80%. Unlike traditional deep learning architecture, the proposed active learning based deep model designed with a smaller number of more localized filters per layer can efficiently learn from small relevant sample batches that can effectively improve performance in the weakly-supervised learning environment and thus is more suitable for several practical applications. An effective dynamic domain adaptive feature weighting scheme can adjust the relative importance of feature dimensions iteratively. Insightful initial feedback gathered from two independent learning modules (a NLP shallow feature based classifier and a deep classifier), modeled to capture complementary information about data characteristics are finally fused together to achieve an impressive 25% average gain in the detection performance.

Data analytics is being increasingly used in cyber-security problems, and found to be useful in cases where data volumes and heterogeneity make it cumbersome for manual assessment by security experts. In practical cyber-security scenarios involving data-driven analytics, obtaining data with annotations (i.e. ground-truth labels) is a challenging and known limiting factor for many supervised security analytics task. Significant portions of the large datasets typically remain unlabelled, as the task of annotation is extensively manual and requires a huge amount of expert intervention. In this paper, we propose an effective active learning approach that can efficiently address this limitation in a practical cyber-security problem of Phishing categorization, whereby we use a human-machine collaborative approach to design a semi-supervised solution. An initial classifier is learnt on a small amount of the annotated data which in an iterative manner, is then gradually updated by shortlisting only relevant samples from the large pool of unlabelled data that are most likely to influence the classifier performance fast. Prioritized Active Learning shows a significant promise to achieve faster convergence in terms of the classification performance in a batch learning framework, and thus requiring even lesser effort for human annotation. An useful feature weight update technique combined with active learning shows promising classification performance for categorizing Phishing/malicious URLs without requiring a large amount of annotated training samples to be available during training. In experiments with several collections of PhishMonger's Targeted Brand dataset, the proposed method shows significant improvement over the baseline by as much as 12%.

This paper describes a method for learning coordination policies in body sensor networks. The learning of a compact coordination policy is important for implementing the policy in sensor nodes with limited memory. We present a novel algorithm, Reinforcement Learning Average Approximation (RLAA), to learn local coordination policies for each sensor node from globally joint rewards. These local policies are obtained by reinforcement learning and averaging state-action tables under a stochastic process model. We show results on a simulation of an existing body sensor network interfaced with transdermal sensors that demonstrate the performance of this learning scheme. Experimental results show that the performance of the RLAA algorithm is significantly better than a random policy and is close to the optimal policy that can be obtained from solving a global Markov Decision Process while the learning step is fast. The results also show that the RLAA algorithm is scalable to networks represented by large state spaces (in terms of number s of sensors and degree of discretization).

The paper describes a method for adaptive transmission of data over error-prone wireless links and suited to the limitations of embedded pervasive sensing systems. The technique performs a multi-resolution wavelet transform on the data followed by interleaving the resulting coefficients among the transmitted packets. The interleaving scheme is designed to facilitate estimation of coefficients lost during transmission by minimizing the correlation between elements in one packet. A correlation model between detail coefficients obtained from a wavelet transform of Gauss-Markov processes is used to estimate the optimal distribution of coefficients to packets. The interleaving ensures that packet loss results in non-contiguous "gaps" in the reconstructed wavelet tree. The original data is reconstructed by polynomial interpolation of the missing coefficients from correlated neighboring coefficient. We present simulation results that show the performance of our scheme on both real-world and simulated datasets.

The paper presents an extension of the particle filtering algorithm that is applicable when an accurate state prediction model cannot be specified but a database of prior state evolution tracks is available. The conventional particle filtering algorithm represents the belief state as a collection of particles, where each particle is a sample from the state space. The particles are updated by applying the state space equations. In the proposed approach, each particle is an instance of a complete state trajectory, drawn from the database of historic state trajectories. An explicit state update model is not required as the trajectory represented by each particle is covers the entire modeling time period. When new observations become available, a proportion of the particles are replaced using trajectories from the database, selected based on distance from the observation. This tracking algorithm is applicable where the state evolves in a complex manner as in the eye of tropical cyclones. The proposed technique is evaluated by tracking selected cyclones from 2005 using a database of cyclone tracks from the previous 25 years.

We present an automated cyclone tracking system that uses images from multiple satellite sources. The system tracks cyclones using infrared images from a Geostationary Operational Environmental Satellite (GOES), precipitation images derived from five satellite sources, and ocean surface wind field satellite images. The system consists of three main components: (i) data preprocessing steps for each data source, (ii) cyclone eye detection algorithms for each data source, and (iii) a filter-based tracker that integrates the eye detection results from each data source. Experimental results show that our prototype system is operationally feasible and has better performance than our prior cyclone tracking system.

Current techniques for cyclone detection and tracking employ NCEP (National Centers for Environmental Prediction) models from in-situ measurements. This solution does not provide true global coverage, unlike remote satellite observations. However it is impractical to use a single Earth orbiting satellite to detect and track events such as cyclones in a continuous manner due to limited spatial and temporal coverage. One solution to alleviate such persistent problems is to utilize heterogeneous sensor data from multiple orbiting satellites. However, this solution requires overcoming other new challenges such as varying spatial and temporal resolution between satellite sensor data, the need to establish correspondence between features from different satellite sensors, and the lack of definitive indicators for cyclone events in some sensor data.

Summary form only given. We discuss a novel framework for integrated sensing and resource management in a distributed intelligent sensor network with sensor nodes whose useful active lifetime is significantly shorter than the required lifetime of the sensor system. Past sensor network research has focused on security and communication; but largely ignored the overall dynamic resource management issue of such distributed systems. Our contribution is in integrated control optimization and resource management algorithms to ensure proper functioning of distributed sensors in extremely limited bandwidth, power, and storage. In this paper, we present a novel genetic algorithm for real-time system control and resource management that handles multiple conflicting objectives and constraints in the distributed system. Our framework is suitable for dynamic environments where the desired system performance and resource usage changes dynamically while being constrained by limited amount of resources.

We discuss a novel control methodology for power management in heterogeneous distributed sensor networks. Many algorithms for resource management in sensor networks require complete knowledge of the external environment and the sensor network system, are rule-based and cannot handle rapidly changing environments; this restricts their use in real-world environments. We present an event based control optimization formulation of the resource management problem and discuss a method to adaptively change desired system performance of the sensor network in response to events. This functionality is critical in field-deployable sensor networks where the available energy is extremely limited. This limitation disallows continuous operation as a very expensive option and necessitates system adaptation as a means to extend operational lifetime in the face of dynamic external events. We show results on synthetic sensor networks where only partially accurate information about the external world and the sensing system is available and illustrate the efficacy of the control algorithm in handling dynamic events with guaranteed minimum system lifespan via efficient usage of energy resources. We show that the control algorithm makes effective control decisions about the use of energy resources with varying sensor reliabilities.

We propose a new machine learning architecture with integrated system control and resource management capability for use in autonomous sensing applications with limited resources. Novel neural network dimensionality reduction with a mixture-of-experts classifier ensures that only relevant information is processed while handling missing sensor data. A genetic optimisation algorithm is used to control the system in the presence of dynamic events, while ensuring that system constraints are met. This tight integration of control optimisation and machine learning algorithms results in a highly efficient intelligent sensor network. The applicability of our technology in remote health monitoring and environmental monitoring is shown.

Dynamic scene perception is very important for autonomous vehicles operating around other moving vehicles and humans. Most work on real-time camera based object tracking from moving platforms has used sparse features or assumed flat scene structures. We have recently extended a real-time, dense stereo system to include real-time, dense optical flow, enabling more comprehensive dynamic scene analysis. We describe algorithms to robustly estimate 6-DOF robot egomotion in the presence of moving objects using dense flow and dense stereo. We then use dense stereo and egomotion estimates to identify other moving objects while the robot itself is moving. We present results showing accurate egomotion estimation and detection of moving people and vehicles under general 6-DOF motion of the robot and independently moving objects. The system runs at 18.3 Hz on a 1.4 GHz Pentium M laptop, computing 160/spl times/120 disparity maps and optical flow fields, egomotion, and moving object segmentation. We believe this is a significant step toward general unconstrained dynamic scene analysis for mobile robots, as well as for improved position estimation where GPS is unavailable.

We discuss a novel control methodology for power management in heterogeneous distributed sensor networks. We present an event based control optimization formulation of the resource management problem and discuss a method to adaptively change the desired system performance of the sensor network in response to events. This functionality is critical in field-deployable sensor networks where continuous operation is expensive and system adaptation is critical for extended operation in the face of dynamic external events. We show results on various synthetic heterogeneous sensor networks where only partially accurate information about the sensing system is available and illustrate the efficacy of the control algorithm in handling such incorrect models with a negligible increase in transmission of the optimal control settings. We show that the run-time performance of the control algorithm scales quadratically with increasing number of sensors.

Dynamic scene perception is currently limited to detection of moving objects from a static platform or scenes with flat backgrounds. We discuss novel real-time methods to segment moving objects in the motion field formed by a moving camera/robotic platform with mostly translational motion. Our solution does not explicitly require any egomotion knowledge, thereby making the solution applicable to mobile outdoor robot problems where no IMU information is available. We address two problems in dynamic scene perception on the move, first using only 2D monocular grayscale images, and second where 3D range information from stereo is also available. Our solution involves real-time optical flow computations, followed by optical flow field preprocessing to highlight moving object boundaries. In the case where range data from stereo is computed, a 3D optical flow field is estimated by combining range information with 2D optical flow estimates, followed by a similar 3D flow field preprocessing step. A segmentation of the flow field using fast flood filling then identifies every moving object in the scene with a unique label. This novel algorithm is expected to be the critical first step in robust recognition of moving vehicles and people from mobile outdoor robots, and therefore offers a robust solution to the problem of dynamic scene perception in the presence of certain kinds of robot motion. It is envisioned that our algorithm will benefit robot scene perception in urban environments for scientific, commercial and defense applications. Results of our real-time algorithm on a mobile robot in a scene with a single moving vehicle are presented.

We discuss techniques to predict the dynamic vehicle response to various natural obstacles. This method can then be used to adjust the vehicle dynamics to optimize performance (e.g. speed) while ensuring that the vehicle is not damaged. This capability opens up a new area of obstacle negotiation for UGVs, where the vehicle moves over certain obstacles, rather than avoiding them, thereby resulting in more effective achievement of objectives. Robust obstacle negotiation and vehicle dynamics prediction requires several key technologies that are discussed in this paper. We detect and segment (label) obstacles using a novel 3D obstacle algorithm. The material of each labelled obstacle (rock, vegetation, etc) is then determined using a texture or color classification scheme. Terrain load-bearing surface models are then constructed using vertical springs to model the compressibility and traversability of each obstacle in front of the vehicle. The terrain model is then combined with the vehicle suspension model to yield an estimate of the maximum safe velocity, and predict the vehicle dynamics as the vehicle follows a path. This end-to-end obstacle negotiation system is envisioned to be useful in optimized path planning and vehicle navigation in terrain conditions cluttered with vegetation, bushes, rocks, etc. Results on natural terrain with various natural materials are presented.

An efficient two-stage algorithm to compute nonlinear features is described. Its implementation on a neural net with adaptive activation functions that raise the input data to an arbitrary power is described. Its use in face recognition with unknown input poses is presented.

A new technique is discussed to recognize human faces under varying aspect views (pose). We first estimate the pose of an unknown human face from a 2D gray-scale image and then transform the unknown face to a reference pose using a feature extraction procedure. A different set of features for discriminating between different individuals are then extracted from these reconstructed faces for recognition. The feature extraction scheme used is known as the maximum representation and discrimination feature method. The advantage of our procedure is that it inherently removes distortions due to pose variations, and therefore requires only single training and/or test face images, which could be at different aspect views. For transformation, it does not require the face to be in the database during training. For recognition, only one aspect view at any pose is necessary.

Texture is a phenomenon in image data that continues to receive attention due to its wide-spread applications, ranging from remotely sensed data, to medical imaging, to military applications. We use a new class of spatial stochastic models called partially ordered Markov models (POMMs) for texture analysis and model selection. POMMs are a generalization of Markov mesh models that have the property that their joint probability density function is an exact, closed form expression in terms of conditional probabilities. Markov random fields (MRFs) do not, in general, have this property. This property of the POMMs has lead to exact and fast computations involving the joint probabilities. We show how these fast algorithms allow POMMs to be used for fitting models to textures, and for supervised texture segmentation. Applications to real data show that the model selection technique gives very good results. POMMs are a broad and general class of models, and have the potential to be applied to diverse areas beyond imaging, such as probabilistic expert systems and artificial intelligence.

Texture is a phenomenon in image data that continues to receive wide-spread interest due to its broad range of applications. The paper focuses on but one of several ways to model textures, namely, the class of stochastic texture models. the authors introduce a new spatial stochastic model called partially ordered Markov models, or POMMs. They show how POMMs are a generalization of a class of models called Markov mesh models, or MMMs, that allow an explicit closed form of the joint probability, just as do MMMs. While POMMs are a type of Markov random field model (MRF), the general MRFs do not have such an explicit closed form of the joint probability. The authors present results on texture synthesis and texture classification, introducing a very fast one-pass texture synthesis algorithm, and show that parameter estimation of natural textures can give quite satisfactory results. They remark that, while the theory underlying POMMs has been applied only to texture analysis, in their most general form, POMMs have the potential to be applied to such diverse areas outside of imaging as probabilistic expert systems, Bayesian hierarchical modeling, influence diagrams, and random graphs and networks.>