Refine
Has Fulltext
- yes (365) (remove)
Year of publication
- 2016 (365) (remove)
Document Type
- Postprint (210)
- Doctoral Thesis (96)
- Article (21)
- Preprint (17)
- Monograph/Edited Volume (8)
- Master's Thesis (4)
- Working Paper (4)
- Habilitation Thesis (2)
- Part of Periodical (2)
- Conference Proceeding (1)
Language
- English (365) (remove)
Keywords
- model (6)
- climate-change (5)
- sentence processing (5)
- German (4)
- aggression (4)
- evolution (4)
- language (4)
- machine learning (4)
- prosody (4)
- protein (4)
Institute
- Mathematisch-Naturwissenschaftliche Fakultät (80)
- Humanwissenschaftliche Fakultät (38)
- Institut für Geowissenschaften (38)
- Institut für Chemie (32)
- Institut für Physik und Astronomie (30)
- Institut für Biochemie und Biologie (24)
- Department Linguistik (23)
- Institut für Mathematik (19)
- Department Psychologie (18)
- Department Sport- und Gesundheitswissenschaften (15)
Computer Security deals with the detection and mitigation of threats to computer networks, data, and computing hardware. This
thesis addresses the following two computer security problems: email spam campaign and malware detection.
Email spam campaigns can easily be generated using popular dissemination tools by specifying simple grammars that serve as message templates. A grammar is disseminated to nodes of a bot net, the nodes create messages by instantiating the grammar at random. Email spam campaigns can encompass huge data volumes and therefore pose a threat to the stability of the infrastructure of email service providers that have to store them. Malware -software that serves a malicious purpose- is affecting web servers, client computers via active content, and client computers through executable files. Without the help of malware detection systems it would be easy for malware creators to collect sensitive information or to infiltrate computers.
The detection of threats -such as email-spam messages, phishing messages, or malware- is an adversarial and therefore intrinsically
difficult problem. Threats vary greatly and evolve over time. The detection of threats based on manually-designed rules is therefore
difficult and requires a constant engineering effort. Machine-learning is a research area that revolves around the analysis of data and the discovery of patterns that describe aspects of the data. Discriminative learning methods extract prediction models from data that are optimized to predict a target attribute as accurately as possible. Machine-learning methods hold the promise of automatically identifying patterns that robustly and accurately detect threats. This thesis focuses on the design and analysis of discriminative learning methods for the two computer-security problems under investigation: email-campaign and malware detection.
The first part of this thesis addresses email-campaign detection. We focus on regular expressions as a syntactic framework, because regular expressions are intuitively comprehensible by security engineers and administrators, and they can be applied as a detection mechanism in an extremely efficient manner. In this setting, a prediction model is provided with exemplary messages from an email-spam campaign. The prediction model has to generate a regular expression that reveals the syntactic pattern that underlies the entire campaign, and that a security engineers finds comprehensible and feels confident enough to use the expression to blacklist further messages at the email server. We model this problem as two-stage learning problem with structured input and output spaces which can be solved using standard cutting plane methods. Therefore we develop an appropriate loss function, and derive a decoder for the resulting optimization problem.
The second part of this thesis deals with the problem of predicting whether a given JavaScript or PHP file is malicious or benign. Recent malware analysis techniques use static or dynamic features, or both. In fully dynamic analysis, the software or script is executed and observed for malicious behavior in a sandbox environment. By contrast, static analysis is based on features that can be extracted directly from the program file. In order to bypass static detection mechanisms, code obfuscation techniques are used to spread a malicious program file in many different syntactic variants. Deobfuscating the code before applying a static classifier can be subjected to mostly static code analysis and can overcome the problem of obfuscated malicious code, but on the other hand increases the computational costs of malware detection by an order of magnitude. In this thesis we present a cascaded architecture in which a classifier first performs a static analysis of the original code and -based on the outcome of this first classification step- the code may be deobfuscated and classified again. We explore several types of features including token $n$-grams, orthogonal sparse bigrams, subroutine-hashings, and syntax-tree features and study the robustness of detection methods and feature types against the evolution of malware over time. The developed tool scans very large file collections quickly and accurately.
Each model is evaluated on real-world data and compared to reference methods. Our approach of inferring regular expressions to filter emails belonging to an email spam campaigns leads to models with a high true-positive rate at a very low false-positive rate that is an order of magnitude lower than that of a commercial content-based filter. Our presented system -REx-SVMshort- is being used by a commercial email service provider and complements content-based and IP-address based filtering.
Our cascaded malware detection system is evaluated on a high-quality data set of almost 400,000 conspicuous PHP files and a collection of more than 1,00,000 JavaScript files. From our case study we can conclude that our system can quickly and accurately process large data collections at a low false-positive rate.
Background: Given the well-established association between perceived stress and quality of life (QoL) in dementia patients and their partners, our goal was to identify whether relationship quality and dyadic coping would operate as mediators between perceived stress and QoL.
Methods: 82 dyads of dementia patients and their spousal caregivers were included in a cross-sectional assessment from a prospective study. QoL was assessed with the Quality of Life in Alzheimer's Disease scale (QoL-AD) for dementia patients and the WHO Quality of Life-BREF for spousal caregivers. Perceived stress was measured with the Perceived Stress Scale (PSS-14). Both partners were assessed with the Dyadic Coping Inventory (DCI). Analyses of correlation as well as regression models including mediator analyses were performed.
Results: We found negative correlations between stress and QoL in both partners (QoL-AD: r = -0.62; p < 0.001; WHO-QOL Overall: r = -0.27; p = 0.02). Spousal caregivers had a significantly lower DCI total score than dementia patients (p < 0.001). Dyadic coping was a significant mediator of the relationship between stress and QoL in spousal caregivers (z = 0.28; p = 0.02), but not in dementia patients. Likewise, relationship quality significantly mediated the relationship between stress and QoL in caregivers only (z = -2.41; p = 0.02).
Conclusions: This study identified dyadic coping as a mediator on the relationship between stress and QoL in (caregiving) partners of dementia patients. In patients, however, we found a direct negative effect of stress on QoL. The findings suggest the importance of stress reducing and dyadic interventions for dementia patients and their partners, respectively.
We present a temperature and fluence dependent Ultrafast X-Ray Diffraction study of a laser-heated antiferromagnetic dysprosium thin film. The loss of antiferromagnetic order is evidenced by a pronounced lattice contraction. We devise a method to determine the energy flow between the phonon and spin system from calibrated Bragg peak positions in thermal equilibrium. Reestablishing the magnetic order is much slower than the cooling of the lattice, especially around the Néel temperature. Despite the pronounced magnetostriction, the transfer of energy from the spin system to the phonons in Dy is slow after the spin-order is lost.
The optical properties of semiconductor nanocrystals (SC NCs) are largely controlled by their size and surface chemistry, i.e., the chemical composition and thickness of inorganic passivation shells and the chemical nature and number of surface ligands as well as the strength of their bonds to surface atoms. The latter is particularly important for CdTe NCs, which – together with alloyed CdxHg1−xTe – are the only SC NCs that can be prepared in water in high quality without the need for an additional inorganic passivation shell. Aiming at a better understanding of the role of stabilizing ligands for the control of the application-relevant fluorescence features of SC NCs, we assessed the influence of two of the most commonly used monodentate thiol ligands, thioglycolic acid (TGA) and mercaptopropionic acid (MPA), on the colloidal stability, photoluminescence (PL) quantum yield (QY), and PL decay behavior of a set of CdTe NC colloids. As an indirect measure for the strength of the coordinative bond of the ligands to SC NC surface atoms, the influence of the pH (pD) and the concentration on the PL properties of these colloids was examined in water and D2O and compared to the results from previous dilution studies with a set of thiol-capped Cd1−xHgxTe SC NCs in D2O. As a prerequisite for these studies, the number of surface ligands was determined photometrically at different steps of purification after SC NC synthesis with Ellman's test. Our results demonstrate ligand control of the pH-dependent PL of these SC NCs, with MPA-stabilized CdTe NCs being less prone to luminescence quenching than TGA-capped ones. For both types of CdTe colloids, ligand desorption is more pronounced in H2O compared to D2O, underlining also the role of hydrogen bonding and solvent molecules.
In the current paradigm of cosmology, the formation of large-scale structures is mainly driven by non-radiating dark matter, making up the dominant part of the matter budget of the Universe. Cosmological observations however, rely on the detection of luminous galaxies, which are biased tracers of the underlying dark matter. In this thesis I present cosmological reconstructions of both, the dark matter density field that forms the cosmic web, and cosmic velocities, for which both aspects of my work are delved into, the theoretical formalism and the results of its applications to cosmological simulations and also to a galaxy redshift survey.The foundation of our method is relying on a statistical approach, in which a given galaxy catalogue is interpreted as a biased realization of the underlying dark matter density field. The inference is computationally performed on a mesh grid by sampling from a probability density function, which describes the joint posterior distribution of matter density and the three dimensional velocity field. The statistical background of our method is described in Chapter ”Implementation of argo”, where the introduction in sampling methods is given, paying special attention to Markov Chain Monte-Carlo techniques. In Chapter ”Phase-Space Reconstructions with N-body Simulations”, I introduce and implement a novel biasing scheme to relate the galaxy number density to the underlying dark matter, which I decompose into a deterministic part, described by a non-linear and scale-dependent analytic expression, and a stochastic part, by presenting a negative binomial (NB) likelihood function that models deviations from Poissonity. Both bias components had already been studied theoretically, but were so far never tested in a reconstruction algorithm. I test these new contributions againstN-body simulations to quantify improvements and show that, compared to state-of-the-art methods, the stochastic bias is inevitable at wave numbers of k≥0.15h Mpc^−1 in the power spectrum in order to obtain unbiased results from the reconstructions. In the second part of Chapter ”Phase-Space Reconstructions with N-body Simulations” I describe and validate our approach to infer the three dimensional cosmic velocity field jointly with the dark matter density. I use linear perturbation theory for the large-scale bulk flows and a dispersion term to model virialized galaxy motions, showing that our method is accurately recovering the real-space positions of the redshift-space distorted galaxies. I analyze the results with the isotropic and also the two-dimensional power spectrum.Finally, in Chapter ”Phase-space Reconstructions with Galaxy Redshift Surveys”, I show how I combine all findings and results and apply the method to the CMASS (for Constant (stellar) Mass) galaxy catalogue of the Baryon Oscillation Spectroscopic Survey (BOSS). I describe how our method is accounting for the observational selection effects inside our reconstruction algorithm. Also, I demonstrate that the renormalization of the prior distribution function is mandatory to account for higher order contributions in the structure formation model, and finally a redshift-dependent bias factor is theoretically motivated and implemented into our method. The various refinements yield unbiased results of the dark matter until scales of k≤0.2 h Mpc^−1in the power spectrum and isotropize the galaxy catalogue down to distances of r∼20h^−1 Mpc in the correlation function. We further test the results of our cosmic velocity field reconstruction by comparing them to a synthetic mock galaxy catalogue, finding a strong correlation between the mock and the reconstructed velocities. The applications of both, the density field without redshift-space distortions, and the velocity reconstructions, are very broad and can be used for improved analyses of the baryonic acoustic oscillations, environmental studies of the cosmic web, the kinematic Sunyaev-Zel’dovic or integrated Sachs-Wolfe effect.
Background:
Skewed body size distributions and the high relative richness of small-bodied taxa are a fundamental
property of a wide range of animal clades. The evolutionary processes responsible for generating these distributions
are well described in vertebrate model systems but have yet to be explored in detail for other major terrestrial
clades. In this study, we explore the macro-evolutionary patterns of body size variation across families of Hexapoda
(insects and their close relatives), using recent advances in phylogenetic understanding, with an aim to investigate
the link between size and diversity within this ancient and highly diverse lineage.
Results:
The maximum, minimum and mean-log body lengths of hexapod families are all approximately log-normally
distributed, consistent with previous studies at lower taxonomic levels, and contrasting with skewed distributions
typical of vertebrate groups. After taking phylogeny and within-tip variation into account, we find no evidence for a
negative relationship between diversification rate and body size, suggesting decoupling of the forces controlling these
two traits. Likelihood-based modeling of the log-mean body size identifies distinct processes operating within
Holometabola and Diptera compared with other hexapod groups, consistent with accelerating rates of size evolution
within these clades, while as a whole, hexapod body size evolution is found to be dominated by neutral processes
including significant phylogenetic conservatism.
Conclusions:
Based on our findings we suggest that the use of models derived from well-studied but atypical clades,
such as vertebrates may lead to misleading conclusions when applied to other major terrestrial lineages. Our results
indicate that within hexapods, and within the limits of current systematic and phylogenetic knowledge, insect
diversification is generally unfettered by size-biased macro-evolutionary processes, and that these processes over large
timescales tend to converge on apparently neutral evolutionary processes. We also identify limitations on available
data within the clade and modeling approaches for the resolution of trees of higher taxa, the resolution of which may
collectively enhance our understanding of this key component of terrestrial ecosystems.
The global carbon cycle is closely linked to Earth’s climate. In the context of continuously unchecked anthropogenic CO₂ emissions, the importance of natural CO₂ bond and carbon storage is increasing. An important biogenic mechanism of natural atmospheric CO₂ drawdown is the photosynthetic carbon fixation in plants and the subsequent longterm deposition of plant detritus in sediments.
The main objective of this thesis is to identify factors that control mobilization and transport of plant organic matter (pOM) through rivers towards sedimentation basins. I investigated this aspect in the eastern Nepalese Arun Valley. The trans-Himalayan Arun River is characterized by a strong elevation gradient (205 − 8848 m asl) that is accompanied by strong changes in ecology and climate ranging from wet tropical conditions in the Himalayan forelad to high alpine tundra on the Tibetan Plateau. Therefore, the Arun is an excellent natural laboratory, allowing the investigation of the effect of vegetation cover, climate, and topography on plant organic matter mobilization and export in tributaries along the gradient.
Based on hydrogen isotope measurements of plant waxes sampled along the Arun River and its tributaries, I first developed a model that allows for an indirect quantification of pOM contributed to the mainsetm by the Arun’s tributaries. In order to determine the role of climatic and topographic parameters of sampled tributary catchments, I looked for significant statistical relations between the amount of tributary pOM export and tributary characteristics (e.g. catchment size, plant cover, annual precipitation or runoff, topographic measures). On one hand, I demonstrated that pOMsourced from the Arun is not uniformly derived from its entire catchment area. On the other, I showed that dense vegetation is a necessary, but not sufficient, criterion for high tributary pOM export. Instead, I identified erosion and rainfall and runoff as key factors controlling pOM sourcing in the Arun Valley. This finding is supported by terrestrial cosmogenic nuclide concentrations measured on river sands along the Arun and its tributaries in order to quantify catchment wide denudation rates. Highest denudation rates corresponded well with maximum pOM mobilization and export also suggesting the link between erosion and pOM sourcing.
The second part of this thesis focusses on the applicability of stable isotope records such as plant wax n-alkanes in sediment archives as qualitative and quantitative proxy for the variability of past Indian Summer Monsoon (ISM) strength. First, I determined how ISM strength affects the hydrogen and oxygen stable isotopic composition (reported as δD and δ18O values vs. Vienna Standard Mean Ocean Water) of precipitation in the Arun Valley and if this amount effect (Dansgaard, 1964) is strong enough to be recorded in potential paleo-ISM isotope proxies. Second, I investigated if potential isotope records across the Arun catchment reflect ISM strength dependent precipitation δD values only, or if the ISM isotope signal is superimposed by winter precipitation or glacial melt. Furthermore, I tested if δD values of plant waxes in fluvial deposits reflect δD values of environmental waters in the respective catchments.
I showed that surface water δD values in the Arun Valley and precipitation δD from south of the Himalaya both changed similarly during two consecutive years (2011 & 2012) with distinct ISM rainfall amounts (~20% less in 2012). In order to evaluate the effect of other water sources (Winter-Westerly precipitation, glacial melt) and evapotranspiration in the Arun Valley, I analysed satellite remote sensing data of rainfall distribution (TRMM 3B42V7), snow cover (MODIS MOD10C1), glacial coverage (GLIMSdatabase, Global Land Ice Measurements from Space), and evapotranspiration (MODIS MOD16A2). In addition to the predominant ISM in the entire catchment I found through stable isotope analysis of surface waters indications for a considerable amount of glacial melt derived from high altitude tributaries and the Tibetan Plateau. Remotely sensed snow cover data revealed that the upper portion of the Arun also receives considerable winter precipitation, but the effect of snow melt on the Arun Valley hydrology could not be evaluated as it takes place in early summer, several months prior to our sampling campaigns. However, I infer that plant wax records and other potential stable isotope proxy archives below the snowline are well-suited for qualitative, and potentially quantitative, reconstructions of past changes of ISM strength.
Plasma carotenoids, tocopherols, and retinol in the age-stratified (35–74 years) general population
(2016)
Blood micronutrient status may change with age. We analyzed plasma carotenoids, α-/γ-tocopherol, and retinol and their associations with age, demographic characteristics, and dietary habits (assessed by a short food frequency questionnaire) in a cross-sectional study of 2118 women and men (age-stratified from 35 to 74 years) of the general population from six European countries. Higher age was associated with lower lycopene and α-/β-carotene and higher β-cryptoxanthin, lutein, zeaxanthin, α-/γ-tocopherol, and retinol levels. Significant correlations with age were observed for lycopene (r = −0.248), α-tocopherol (r = 0.208), α-carotene (r = −0.112), and β-cryptoxanthin (r = 0.125; all p < 0.001). Age was inversely associated with lycopene (−6.5% per five-year age increase) and this association remained in the multiple regression model with the significant predictors (covariables) being country, season, cholesterol, gender, smoking status, body mass index (BMI (kg/m2)), and dietary habits. The positive association of α-tocopherol with age remained when all covariates including cholesterol and use of vitamin supplements were included (1.7% vs. 2.4% per five-year age increase). The association of higher β-cryptoxanthin with higher age was no longer statistically significant after adjustment for fruit consumption, whereas the inverse association of α-carotene with age remained in the fully adjusted multivariable model (−4.8% vs. −3.8% per five-year age increase). We conclude from our study that age is an independent predictor of plasma lycopene, α-tocopherol, and α-carotene.
Portal Wissen = Point
(2016)
A point is more than meets the eye. In geometry, a point is an object with zero dimensions – it is there but takes up little space. You may assume that something so small is easily overlooked. A closer look reveals that points are everywhere and play a significant role in many areas. In physics, for example, a mass point is the highest possible idealization of a body, which is the theoretical notion that the entire mass of a body is concentrated in a point, its “center of mass”.
Points are at the beginning (starting points), at intersections (pivot points), and at the end (final points). A point symbolizes great precision. There is a reason we “get to the point”. In writing, a point abbreviates, structures, and finalizes what is said. Physicians puncture, and athletes collect points on playing fields, courses, and on tables.
It’s no wonder that researchers are “surrounded” by points and work with them every day: Points bring order to chaos, structure the unexplained, and name the nameless. A point is often the beginning, an entry to worlds, findings, or problems.
Points are for everyone, though. German mathematician Oskar Perron wrote, “A point is exactly what the intelligent yet innocent, uncorrupted reader imagines it to be.” We want to follow up on this quotation: The latest edition of Portal Wissen offers exciting starting points, analyzes points of view, and gets right to the point.
We follow a physicist to the sun – the center point of our solar system – to ponder the origin of solar eruptions. We talked to a marketing professor about turning contentious points into successful deals during negotiation. Business information experts present leverage points that prepare both humans and machines for factories in the age of Industry 4.0. Enthusiastic entrepreneurs show us how their research became the starting point of a successful business idea – and also make the world a bit better. Geoscientists explain why the weather phenomenon El Niño causes – wet and dry – flashpoints. Just to name a few of many points …
We hope our magazine scores points with you and wish you an inspiring read!
The Editorial
Portal Wissen = small
(2016)
Let’s be honest: even science wants to make it big, at least when it comes to discovering new knowledge. Yet if one thing belongs in the annals of successful research, it is definitely small things. Scientists have long understood that their job is to explore things that they don’t see right away. Seneca once wrote, “If something is smaller than the great, this does not mean at all that it is insignificant.”
The smallest units of life, such as bacteria or viruses, can often have powerful effects. And again and again, (seemingly) large things must first be disassembled or reduced to small pieces in order to recognize their nature. One of the greatest secrets of our world – the atom, the smallest, if no longer indivisible, unit of chemical elements – revealed itself only by looking at its diminutive size. By no means is ‘small’ (German: klein) merely a counterpoint to large, at least in linguistic terms; the word comes from West Germanic klaini, which means ‘fine’ or ‘delicate,’ and is also related to the English word ‘clean.’ Fine and clean – certainly something worth striving for in scientific work. And a bit of attention to detail doesn’t hurt either.
This doesn’t mean that researchers can be smallminded; they should be ready to expect the unexpected and to adjust their work accordingly. And even if they cannot attain their goals in the short term, they need staying power to keep themselves from being talked down, from giving up.
Strictly speaking, research is like putting together a puzzle with tons of tiny pieces; you don’t want it to end. Every discovery worthy of a Nobel Prize, every major research project, has to start with a small idea, with a tiny spark, and then the planning of the minutest details can begin. What follows is work focused on minuscule details: hours of interviews searching for the secret of the cerebellum (Latin for ‘little brain’), days of field studies searching for Lilliputian forms of life, weeks of experimentation meant to render visible the microscopically tiny, months of archival research that brings odds and ends to light, or years of reading fine print. All while hunting for a big hit...
This is why we’ve assembled a few ‘little’ stories about research at the University of Potsdam, under the motto: small, but look out! Nutritional scientists are working on rescuing some of the earth’s smaller residents – mice – from the fate of ‘lab rats’ by developing alternatives to animal testing. Linguists are using innovative methods in several projects to investigate how small children learn languages. Astrophysicists in Potsdam are scanning the skies above Babelsberg for the billions of stars in the Magellan Cloud, which only seem tiny from down here. The Research Center Sanssouci, initiated by the Prussian Palaces and Gardens Foundation and the University of Potsdam, is starting small but will bring about great things for Potsdam’s cultural landscape. Biologists are drilling down to the smallest building blocks of life, looking for genes in barley so that new strains with positive characteristics can be cultivated.
Like we said: little things. Have fun reading!
The Editorial