Community sense and response systems

The proliferation of smartphones and other internet-enabled, sensor-equipped consumer devices enables us to sense and act upon the physical environment in unprecedented ways. This thesis considers Community Sense-and-Response (CSR) systems, a new class of web application for acting on sensory data gathered from participants' personal smart devices. The thesis describes how rare events can be reliably detected using a decentralized anomaly detection architecture that performs client-side anomaly detection and server-side event detection. After analyzing this decentralized anomaly detection approach, the thesis describes how weak but spatially structured events can be detected, despite significant noise, when the events have a sparse representation in an alternative basis. Finally, the thesis describes how the statistical models needed for client-side anomaly detection may be learned efficiently, using limited space, via coresets.

The Caltech Community Seismic Network (CSN) is a prototypical example of a CSR system that harnesses accelerometers in volunteers' smartphones and consumer electronics. Using CSN, this thesis presents the systems and algorithmic techniques to design, build and evaluate a scalable network for real-time awareness of spatial phenomena such as dangerous earthquakes.

Detection of specific bacteria in water: implications of survival strategy

It is widely recognised that conventional culture techniques may underestimate true viable bacterial numbers by several orders of magnitude. The basis of this discrepancy is that a culture in or on media of high nutrient concentration is highly selective (either through ”nutrient shock” or failure to provide vital co-factors) and decreases apparent diversity; thus it is unrepresentative of the natural community. In addition, the non-culturable but viable state (NCBV) is a strategy adopted by some bacteria as a response to environmental stress. The basis for the non-culturable state is that cells placed in conditions present in the environment cannot be recultured but can be shown to maintain their viability. Consequently, these cells would not be detected by standard water quality techniques that are based on culture. In the case of pathogens, it may explain outbreaks of disease in populations that have not come into contact with the pathogen. However, the NCBV state is difficult to attribute, due to the failure to distinguish between NCBV and non-viable cells. This article will describe experiences with the fish pathogen Aeromonas salmonicida subsp. salmonicida and the application of molecular techniques for its detection and physiological analysis.

Soil microbial community analysis using denaturing gradient gel electrophoresis

The most biological diversity on this planet is probably harbored in soils. Understanding the diversity and function of the microbiological component of soil poses great challenges that are being overcome by the application of molecular biological approaches. This review covers one of many approaches being used: separation of polymerase chain reaction (PCR) amplicons using denaturing gradient gel electrophoresis (DGGE). Extraction of nucleic acids directly from soils allows the examination of a community without the limitation posed by cultivation. Polymerase chain reaction provides a means to increase the numbers of a target for its detection on gels. Using the rRNA genes as a target for PCR provides phylogenetic information on populations comprising communities. Fingerprints produced by this method have allowed spatial and temporal comparisons of soil communities within and between locations or among treatments. Numerous samples can be compared because of the rapid high throughput nature of this method. Scientists now have the means to begin addressing complex ecological questions about the spatial, temporal, and nutritional interactions faced by microbes in the soil environment.

Design and implementation of robust systems for secure malware detection

Malicious software (malware) have significantly increased in terms of number and effectiveness during the past years. Until 2006, such software were mostly used to disrupt network infrastructures or to show coders’ skills. Nowadays, malware constitute a very important source of economical profit, and are very difficult to detect. Thousands of novel variants are released every day, and modern obfuscation techniques are used to ensure that signature-based anti-malware systems are not able to detect such threats. This tendency has also appeared on mobile devices, with Android being the most targeted platform. To counteract this phenomenon, a lot of approaches have been developed by the scientific community that attempt to increase the resilience of anti-malware systems. Most of these approaches rely on machine learning, and have become very popular also in commercial applications. However, attackers are now knowledgeable about these systems, and have started preparing their countermeasures. This has lead to an arms race between attackers and developers. Novel systems are progressively built to tackle the attacks that get more and more sophisticated. For this reason, a necessity grows for the developers to anticipate the attackers’ moves. This means that defense systems should be built proactively, i.e., by introducing some security design principles in their development. The main goal of this work is showing that such proactive approach can be employed on a number of case studies. To do so, I adopted a global methodology that can be divided in two steps. First, understanding what are the vulnerabilities of current state-of-the-art systems (this anticipates the attacker’s moves). Then, developing novel systems that are robust to these attacks, or suggesting research guidelines with which current systems can be improved. This work presents two main case studies, concerning the detection of PDF and Android malware. The idea is showing that a proactive approach can be applied both on the X86 and mobile world. The contributions provided on this two case studies are multifolded. With respect to PDF files, I first develop novel attacks that can empirically and optimally evade current state-of-the-art detectors. Then, I propose possible solutions with which it is possible to increase the robustness of such detectors against known and novel attacks. With respect to the Android case study, I first show how current signature-based tools and academically developed systems are weak against empirical obfuscation attacks, which can be easily employed without particular knowledge of the targeted systems. Then, I examine a possible strategy to build a machine learning detector that is robust against both empirical obfuscation and optimal attacks. Finally, I will show how proactive approaches can be also employed to develop systems that are not aimed at detecting malware, such as mobile fingerprinting systems. In particular, I propose a methodology to build a powerful mobile fingerprinting system, and examine possible attacks with which users might be able to evade it, thus preserving their privacy. To provide the aforementioned contributions, I co-developed (with the cooperation of the researchers at PRALab and Ruhr-Universität Bochum) various systems: a library to perform optimal attacks against machine learning systems (AdversariaLib), a framework for automatically obfuscating Android applications, a system to the robust detection of Javascript malware inside PDF files (LuxOR), a robust machine learning system to the detection of Android malware, and a system to fingerprint mobile devices. I also contributed to develop Android PRAGuard, a dataset containing a lot of empirical obfuscation attacks against the Android platform. Finally, I entirely developed Slayer NEO, an evolution of a previous system to the detection of PDF malware. The results attained by using the aforementioned tools show that it is possible to proactively build systems that predict possible evasion attacks. This suggests that a proactive approach is crucial to build systems that provide concrete security against general and evasion attacks.

Accuracy of proactive case finding for mental disorders by community informants in Nepal.

BACKGROUND: Accurate detection of persons in need of mental healthcare is crucial to reduce the treatment gap between psychiatric burden and service use in low- and middle-income (LAMI) countries. AIMS: To evaluate the accuracy of a community-based proactive case-finding strategy (Community Informant Detection Tool, CIDT), involving pictorial vignettes, designed to initiate pathways for mental health treatment in primary care settings. METHOD: Community informants using the CIDT identified screen positive (n = 110) and negative persons (n = 85). Participants were then administered the Composite International Diagnostic Interview (CIDI). RESULTS: The CIDT has a positive predictive value of 0.64 (0.68 for adults only) and a negative predictive value of 0.93 (0.91 for adults only). CONCLUSIONS: The CIDT has promising detection properties for psychiatric caseness. Further research should investigate its potential to increase demand for, and access to, mental health services.

Pneumococcal aetiology and serotype distribution in paediatric community-acquired pneumonia.

Community-acquired pneumonia (CAP) is a major cause of morbidity in children. This study estimated the proportion of children with pneumococcal CAP among children hospitalised with CAP in Belgium and describes the causative serotype distribution after implementation of the 7-valent pneumococcal conjugate vaccine. Children 0-14 years hospitalised with X-ray-confirmed CAP were prospectively enrolled in a multicentre observational study. Acute and convalescent blood samples were collected. Pneumococcal aetiology was assessed by conventional methods (blood or pleural fluid cultures with Quellung reaction capsular typing or polymerase chain reaction [PCR] in pleural fluid), and recently developed methods (real-time PCR in blood and World Health Organization-validated serotype-specific serology). A total of 561 children were enrolled. Pneumococcal aetiology was assessed by conventional methods in 539, serology in 171, and real-time PCR in blood in 154. Pneumococcal aetiology was identified in 12.2% (66/539) of the children by conventional methods alone but in 73.9% by the combination of conventional and recently developed methods. The pneumococcal detection rate adjusted for the whole study population was 61.7%. Serotypes 1 (42.3%), 5 (16.0%), and 7F(7A) (12.8%) were predominant. In conclusion, Streptococcus pneumoniae remains the predominant bacteria in children hospitalised for CAP in Belgium after implementation of 7-valent pneumococcal conjugate vaccine, with non-vaccine-serotypes accounting for the majority of cases. The use of recently developed methods improves diagnosis of pneumococcal aetiology.

Theoretical analysis of ocean color radiances anomalies and implications for phytoplankton groups detection in case 1 waters

Past years have seen the development of different approaches to detect phytoplankton groups from space. One of these methods, the PHYSAT one, is empirically based on reflectance anomalies. Despite observations in good agreement with in situ measurements, the underlying theoretical explanation of the method is still missing and needed by the ocean color community as it prevents improvements of the methods and characterization of uncertainties on the inversed products. In this study, radiative transfer simulations are used in addition to in situ measurements to understand the organization of the signals used in PHYSAT. Sensitivity analyses are performed to assess the impact of the variability of the following three parameters on the reflectance anomalies: specific phytoplankton absorption, colored dissolved organic matter absorption, and particles backscattering. While the later parameter explains the largest part of the anomalies variability, results show that each group is generally associated with a specific bio-optical environment which should be considered to improve methods of phytoplankton groups detection.

High-resolution view of the spring bloom initiation and net community production in the Subantarctic Southern Ocean using glider data

In the Southern Ocean, there is increasing evidence that seasonal to subseasonal temporal scales, and meso- to submesoscales play an important role in understanding the sensitivity of ocean primary productivity to climate change. This drives the need for a high-resolution approach to re- solving biogeochemical processes. In this study, 5.5 months of continuous, high-resolution (3 h, 2 km horizontal resolution) glider data from spring to summer in the Atlantic Subantarctic Zone is used to investigate: (i) the mechanisms that drive bloom initiation and high growth rates in the region and (ii) the seasonal evolution of water column production and respiration. Bloom initiation dates were analysed in the context of upper ocean boundary layer physics highlighting sensitivities of different bloom detection methods to different environmental processes. Model results show that in early spring (September to mid-November) increased rates of net community production (NCP) are strongly affected by meso- to submesoscale features. In late spring/early summer (late-November to mid-December) seasonal shoaling of the mixed layer drives a more spatially homogenous bloom with maximum rates of NCP and chlorophyll biomass. A comparison of biomass accumulation rates with a study in the North Atlantic highlights the sensitivity of phytoplankton growth to fine-scale dynamics and emphasizes the need to sample the ocean at high resolution to accurately resolve phytoplankton phenology and improve our ability to estimate the sensitivity of the biological carbon pump to climate change.

Use of culture and molecular analysis to determine the effect of antibiotic treatment on microbial community diversity and abundance during exacerbation in patients with cystic fibrosis

Background: Anaerobic bacteria are increasingly regarded as important in cystic fibrosis (CF) pulmonary infection. The aim of this study was to determine the effect of antibiotic treatment on aerobic and anaerobic microbial community diversity and abundance during exacerbations in patients with CF.

Methods: Sputum was collected at the start and completion of antibiotic treatment of exacerbations and when clinically stable. Bacteria were quantified and identified following culture, and community composition was also examined using culture-independent methods.

Results: Pseudomonas aeruginosa or Burkholderia cepacia complex were detected by culture in 24/26 samples at the start of treatment, 22/26 samples at completion of treatment and 11/13 stable samples. Anaerobic bacteria were detected in all start of treatment and stable samples and in 23/26 completion of treatment samples. Molecular analysis showed greater bacterial diversity within sputum samples than was detected by culture; there was reasonably good agreement between the methods for the presence or absence of aerobic bacteria such as P aeruginosa (kappa=0.74) and B cepacia complex (kappa=0.92), but agreement was poorer for anaerobes. Both methods showed that the composition of the bacterial community varied between patients but remained relatively stable in most individuals despite treatment. Bacterial abundance decreased transiently following treatment, with this effect more evident for aerobes (median decrease in total viable count 2.3 x 10(7) cfu/g, p=0.005) than for anaerobes (median decrease in total viable count 3 x 10(6) cfu/g, p=0.046).

Conclusion: Antibiotic treatment targeted against aerobes had a minimal effect on abundance of anaerobes and community composition, with both culture and molecular detection methods required for comprehensive characterisation of the microbial community in the CF lung. Further studies are required to determine the clinical significance of and optimal treatment for these newly identified bacteria.

Accuracy of LightCycler® SeptiFast for the detection and identification of pathogens in the blood of patients with suspected sepsis: a systematic review protocol

Background: There is growing interest in the potential utility of molecular diagnostics in improving the detection of life-threatening infection (sepsis). LightCycler® SeptiFast is a multipathogen probebased real-time PCR system targeting DNA sequences of bacteria and fungi present in blood samples within a few hours. We report here the protocol of the first systematic review of published clinical diagnostic accuracy studies of this technology when compared with blood culture in the setting of suspected sepsis. Methods/design: Data sources: the Cochrane Database of Systematic Reviews, the Database of Abstracts of Reviews of Effects (DARE), the Health Technology Assessment Database (HTA), the NHS Economic Evaluation Database (NHSEED), The Cochrane Library, MEDLINE, EMBASE, ISI Web of Science, BIOSIS Previews, MEDION and the Aggressive Research Intelligence Facility Database (ARIF). Study selection: diagnostic accuracy studies that compare the real-time PCR technology with standard culture results performed on a patient's blood sample during the management of sepsis. Data extraction: three reviewers, working independently, will determine the level of evidence, methodological quality and a standard data set relating to demographics and diagnostic accuracy metrics for each study. Statistical analysis/data synthesis: heterogeneity of studies will be investigated using a coupled forest plot of sensitivity and specificity and a scatter plot in Receiver Operator Characteristic (ROC) space. Bivariate model method will be used to estimate summary sensitivity and specificity. The authors will investigate reporting biases using funnel plots based on effective sample size and regression tests of asymmetry. Subgroup analyses are planned for adults, children and infection setting (hospital vs community) if sufficient data are uncovered. Dissemination: Recommendations will be made to the Department of Health (as part of an open-access HTA report) as to whether the real-time PCR technology has sufficient clinical diagnostic accuracy potential to move forward to efficacy testing during the provision of routine clinical care.

MICROBIAL COMMUNITY COMPOSITION IN THE LUNGS OF PATIENTS WITH CYSTIC FIBROSIS AND NON-CF BRONCHIECTASIS

Introduction and Aims: Persistent bacterial infection is a major cause of morbidity and mortality in patients with both Cystic Fibrosis (CF) and non-CF Bronchiectasis (non-CFBX). Numerous studies have shown that CF and non-CFBX airways are colonised by a complex microbiota. However, many bacteria are difficult, if not impossible, to culture by conventional laboratory techniques. Therefore, molecular detection techniques offer a more comprehensive view of bacterial diversity within clinical specimens. The objective of this study was to characterise and compare bacterial diversity and relative abundance in patients with CF and non-CFBX during exacerbation and when clinically stable.

Methods: Sputum samples were collected from CF (n=50 samples) and non-CFBX (n=52 samples) patients at the start and end of treatment for an infective exacerbation and when clinically stable. Pyrosequencing was used to assess the microbial diversity and relative genera (or the closest possibly taxonomic order) abundance within the samples. Each sequence read was defined based on 3% difference.

Results: High-throughput pyrosequencing allowed a sensitive and detailed examination of microbial community composition. Rich microbial communities were apparent within both CF (171 species-level phylotypes per genus) and non-CFBX airways (144 species-level phylotypes per genus). Relative species distribution within those two environments was considerably different; however, relatively few genera formed a core of microorganisms, representing approximately 90% of all sequences, which dominated both environments. Relative abundance based on observed operational taxonomic units demonstrated that the most abundant bacteria in CF were Pseudomonas (28%), Burkholderia (22%), Streptococcus (13%), family Pseudomonadaceae (8%) and Prevotella (6%). In contrast, the most commonly detected operational taxonomic units in non-CFBX were Haemophilus (22%), Streptococcus (14%), other (unassigned taxa) (11%), Pseudomonas (10%), Veillonella (7%) and Prevotella (6%).

Conclusions: These results suggest that distinctive microbial communities are associated with infection and/or colonisation in patients with both CF and non-CFBX. Although relatively high species richness was observed within the two environments, each was dominated by different core taxa. This suggests that differences in the lung environment of these two diseases may affect adaptability of the relevant bacterial taxa.

Sarcasm Detection on Czech and English Twitter

This paper presents a machine learning approach to sarcasm detection on Twitter in two languages – English and Czech. Although there has been some research in sarcasm detection in languages other than English (e.g., Dutch, Italian, and Brazilian Portuguese), our work is the first attempt at sarcasm detection in the Czech language. We created a large Czech Twitter corpus consisting of 7,000 manually-labeled tweets and provide it to the community. We evaluate two classifiers with various combinations of features on both the Czech and English datasets. Furthermore, we tackle the issues of rich Czech morphology by examining different preprocessing techniques. Experiments show that our language-independent approach significantly outperforms adapted state-of-the-art methods in English (F-measure 0.947) and also represents a strong baseline for further research in Czech (F-measure 0.582).

Anomaly Detection for Data with Spatial Attributes

The problem of detecting spatially-coherent groups of data that exhibit anomalous behavior has started to attract attention due to applications across areas such as epidemic analysis and weather forecasting. Earlier efforts from the data mining community have largely focused on finding outliers, individual data objects that display deviant behavior. Such point-based methods are not easy to extend to find groups of data that exhibit anomalous behavior. Scan Statistics are methods from the statistics community that have considered the problem of identifying regions where data objects exhibit a behavior that is atypical of the general dataset. The spatial scan statistic and methods that build upon it mostly adopt the framework of defining a character for regions (e.g., circular or elliptical) of objects and repeatedly sampling regions of such character followed by applying a statistical test for anomaly detection. In the past decade, there have been efforts from the statistics community to enhance efficiency of scan statstics as well as to enable discovery of arbitrarily shaped anomalous regions. On the other hand, the data mining community has started to look at determining anomalous regions that have behavior divergent from their neighborhood.In this chapter,we survey the space of techniques for detecting anomalous regions on spatial data from across the data mining and statistics communities while outlining connections to well-studied problems in clustering and image segmentation. We analyze the techniques systematically by categorizing them appropriately to provide a structured birds eye view of the work on anomalous region detection;we hope that this would encourage better cross-pollination of ideas across communities to help advance the frontier in anomaly detection.

Habitat-species interactions in a carnivore community

Na maioria das area protegidas nacionais existe uma considerável falta de informação científica relativa aos mamíferos carnívoros. A Reserva Natural da Serra da Malcata, localizada no centro-este de Portugal, tem vindo a desenvolver, desde os últimos 20 anos, estudos de ecologia e esta tese pretende dar seguimento a esse esforço de monitorização, através do desenvolvimento de métodos simples e eficientes, de monitorização de carnívoros, que possam servir como percursores de trabalhos a longo-prazo em áreas relevantes para a conservação. Através do uso de armadilhagem fotográfica, foram estudadas relações espécies-habitats para 5 espécies: gato-bravo (Felis silvestris), fuinha (Martes foina), raposa (Vulpes vulpes), gineta (Genetta genetta), e sacarrabos (Herpestes ichneumon). Foram desenvolvidos métodos para se determinar a densidade absoluta de raposa e gineta e a população de gato-bravo foi estudada em detalhe. As principais conclusões do estudo foram: 1) a ocupação de raposas é uniforme e parece ser independente de variáveis ambientais; 2) a ocupação de fuinha encontra-se relacionada com variáveis de habitat, estrutura paisagística e presas; 3) a ocupação de gineta está relacionada com a cobertura de folhosas e distribuição de presas; 4) para o sacarrabos verificase que a ocupação é influenciada pelas extensões de habitat arbustivos, 5) a população de gato-bravo sofreu um forte declínio durante o trabalho e requer urgentes medidas de conservação. Metodologicamente foi demonstrada a importância da modelação das probabilidades de detecção para espécies para as quais este parâmetros apresenta valores baixos ou é muito variável. Esta tese também demonstrou a grande importância da Serra da Malcata para a conservação de carnívoros e a necessidade do desenvolvimento de técnicas de monitorização padronizadas para uma correcta gestão adaptativa das áreas protegidas.

Advances in therapeutic risk management through signal detection and risk minimisation tool analyses

Les quatre principales activités de la gestion de risque thérapeutique comportent l’identification, l’évaluation, la minimisation, et la communication du risque. Ce mémoire aborde les problématiques liées à l’identification et à la minimisation du risque par la réalisation de deux études dont les objectifs sont de: 1) Développer et valider un outil de « data mining » pour la détection des signaux à partir des banques de données de soins de santé du Québec; 2) Effectuer une revue systématique afin de caractériser les interventions de minimisation de risque (IMR) ayant été implantées. L’outil de détection de signaux repose sur la méthode analytique du quotient séquentiel de probabilité (MaxSPRT) en utilisant des données de médicaments délivrés et de soins médicaux recueillis dans une cohorte rétrospective de 87 389 personnes âgées vivant à domicile et membres du régime d’assurance maladie du Québec entre les années 2000 et 2009. Quatre associations « médicament-événement indésirable (EI) » connues et deux contrôles « négatifs » ont été utilisés. La revue systématique a été faite à partir d’une revue de la littérature ainsi que des sites web de six principales agences réglementaires. La nature des RMIs ont été décrites et des lacunes de leur implémentation ont été soulevées. La méthode analytique a mené à la détection de signaux dans l'une des quatre combinaisons médicament-EI. Les principales contributions sont: a) Le premier outil de détection de signaux à partir des banques de données administratives canadiennes; b) Contributions méthodologiques par la prise en compte de l'effet de déplétion des sujets à risque et le contrôle pour l'état de santé du patient. La revue a identifié 119 IMRs dans la littérature et 1,112 IMRs dans les sites web des agences réglementaires. La revue a démontré qu’il existe une augmentation des IMRs depuis l’introduction des guides réglementaires en 2005 mais leur efficacité demeure peu démontrée.