Refine
H-BRS Bibliography
- no (60) (remove)
Departments, institutes and facilities
Document Type
- Part of a Book (27)
- Conference Object (19)
- Article (8)
- Book (monograph, edited volume) (3)
- Patent (2)
- Contribution to a Periodical (1)
Year of publication
- 2018 (60) (remove)
Keywords
Towards explaining deep learning networks to distinguish facial expressions of pain and emotions
(2018)
Deep learning networks are successfully used for object and face recognition in images and videos. In order to be able to apply such networks in practice, for example in hospitals as a pain recognition tool, the current procedures are only suitable to a limited extent. The advantage of deep learning methods is that they can learn complex non-linear relationships between raw data and target classes without limiting themselves to a set of hand-crafted features provided by humans. However, the disadvantage is that due to the complexity of these networks, it is not possible to interpret the knowledge that is stored inside the network. It is a black-box learning procedure. Explainable Artificial Intelligence (AI) approaches mitigate this problem by extracting explanations for decisions and representing them in a human-interpretable form. The aim of this paper is to investigate the explainable AI method Layer-wise Relevance Propagation (LRP) and apply it to explain how a deep learning network distinguishes facial expressions of pain from facial expressions of emotions such as happiness and disgust.
This paper describes a dynamic, model-based approach for estimating intensities of 22 out of 44 different basic facial muscle movements. These movements are defined as Action Units (AU) in the Facial Action Coding System (FACS) [1]. The maximum facial shape deformations that can be caused by the 22 AUs are represented as vectors in an anatomically based, deformable, point-based face model. The amount of deformation along these vectors represent the AU intensities, and its valid range is [0, 1]. An Extended Kalman Filter (EKF) with state constraints is used to estimate the AU intensities. The focus of this paper is on the modeling of constraints in order to impose the anatomically valid AU intensity range of [0, 1]. Two process models are considered, namely constant velocity and driven mass-spring-damper. The results show the temporal smoothing and disambiguation effect of the constrained EKF approach, when compared to the frame-by-frame model fitting approach ‘Regularized Landmark Mean-Shift (RLMS)’ [2]. This effect led to more than 35% increase in performance on a database of posed facial expressions.
Cholinergic polymodal chemosensory cells in the mammalian urethra (urethral brush cells = UBC) functionally express the canonical bitter and umami taste transduction signaling cascade. Here, we aimed to determine whether UBC are functionally equipped for the perception of salt through ENaC (epithelial sodium channel). Cholinergic UBC were isolated from ChAT-eGFP reporter mice (ChAT = choline acetyltransferase). RT-PCR showed mRNA expression of ENaC subunits Scnn1a, Scnn1b, and Scnn1g in urethral epithelium and isolated UBC. Scnn1a could also be detected by next generation sequencing in 4/6 (66%) single UBC, two of them also expressed the bitter receptor Tas2R108. Strong expression of Scnn1a was seen in some urothelial umbrella cells and in 65% of UBC (30/46 cells) in a Scnn1a reporter mouse strain. Intracellular [Ca2+] was recorded in isolated UBC stimulated with the bitter substance denatonium benzoate (25 mM), ATP (0.5 mM) and NaCl (50 mM, on top of 145 mM Na+ and 153 mM Cl- baseline in buffer); mannitol (150 mM) served as osmolarity control. NaCl, but not mannitol, evoked an increase in intracellular [Ca2+] in 70% of the tested UBC. The NaCl-induced effect was blocked by the ENaC inhibitor amiloride (IC50 = 0.471 mu M). When responses to both NaCl and denatonium were tested, all three possible positive response patterns occurred in a balanced distribution: 42% NaCl only, 33% denatonium only, 25% to both stimuli. A similar reaction pattern was observed with ATP and NaCl as test stimuli. About 22% of the UBC reacted to all three stimuli. Thus, NaCl evokes calcium responses in several UBC, likely involving an amiloride-sensitive channel containing alpha-ENaC. This feature does not define a new subpopulation of UBC, but rather emphasizes their polymodal character. The actual function of alpha-ENaC in cholinergic UBC-salt perception, homeostatic ion transport, mechanoreception-remains to be determined.
Evolutionary conservation of the antimicrobial function of mucus: a first defence against infection
(2018)
Mucus layers often provide a unique and multi-functional hydrogel interface between the epithelial cells of organisms and their external environment. Mucus has exceptional properties including elasticity, changeable rheology and an ability to self-repair by reannealing, and is therefore an ideal medium for trapping and immobilising pathogens and serving as a barrier to microbial infection. The ability to produce a functional surface mucosa was an important evolutionary step, which evolved first in the Cnidaria, which includes corals, and the Ctenophora. This allowed the exclusion of non-commensal microbes and the subsequent development of the mucus-lined digestive cavity seen in higher metazoans. The fundamental architecture of the constituent glycoprotein mucins is also evolutionarily conserved. Although an understanding of the biochemical interactions between bacteria and the mucus layer are important to the goal of developing new antimicrobial strategies, they remain relatively poorly understood. This review summarises the physicochemical properties and evolutionary importance of mucus, which make it so successful in the prevention of bacterial infection. In addition, the strategies developed by bacteria to counteract the mucus layer are also explored.
The epithelial sodium channel (ENaC) is a critical regulator of vertebrate electrolyte homeostasis. ENaC is the only constitutively open ion channel in the degenerin/ENaC protein family, and its expression, membrane abundance, and open probability therefore are tightly controlled. The canonical ENaC is composed of three subunits (, , and ), but a fourth -subunit may replace and form atypical -ENaCs. Using Xenopus laevis as a model, here we found that mRNAs of the - and -subunits are differentially expressed in different tissues and that -ENaC predominantly is present in the urogenital tract. Using whole-cell and single-channel electrophysiology of oocytes expressing Xenopus - or -ENaC, we demonstrate that the presence of the -subunit enhances the amount of current generated by ENaC due to an increased open probability, but also changes current into a transient form. Activity of canonical ENaCs is critically dependent on proteolytic processing of the - and -subunits, and immunoblotting with epitope-tagged ENaC subunits indicated that, unlike -ENaC, the -subunit does not undergo proteolytic maturation by the endogenous protease furin. Furthermore, currents generated by -ENaC were insensitive to activation by extracellular chymotrypsin, and presence of the -subunit prevented cleavage of -ENaC at the cell surface. Our findings suggest that subunit composition constitutes an additional level of ENaC regulation, and we propose that the Xenopus -ENaC subunit represents a functional example that demonstrates the importance of proteolytic maturation during ENaC evolution.
In Fortführung zu den drei erfolgreichen „Usable Security und Privacy“ Workshops der letzten drei Jahre, sollen in einem vierten ganztätigen wissenschaftlichen Workshop auf der diesjährigen Mensch und Computer sechs bis acht Arbeiten auf dem Gebiet Usable Security and Privacy vorgestellt und diskutiert werden. Vorgesehen sind Beiträge aus Forschung und Praxis, die neue nutzerzentrierte Ansätze aber auch praxisrelevante Lösungen zur nutzerzentrierten Entwicklung und Ausgestaltung von digitalen Schutzmechanismen thematisieren. Mit dem Workshop soll das etablierte Forum weiterentwickelt werden, in dem sich Experten aus unterschiedlichen Domänen, z. B. dem Usability-Engineering und Security-Engineering, transdisziplinär austauschen können. Der Workshop wird von den Organisatoren als klassischer wissenschaftlicher Workshop ausgestaltet. Ein Programmkomitee bewertet die Einreichungen und wählt daraus die zur Präsentation akzeptierten Beiträge aus. Diese werden zudem im Poster- und Workshopband der Mensch und Computer 2018 veröffentlicht.
Kontemporäre Service-orientierte Systeme sind hochgradig vernetzt und haben zudem die Eigenschaft massiv-skalierbar zu sein. Diese Charakteristiken stellen im besonderen Maße Anforderungen an die Datensicherheit der Anwender solcher Systeme und damit primär an alle Stakeholder der Softwareentwicklung, die in der Verantwortung sind, passgenaue Sicherheitsmechanismen effektiv in die Softwareprodukte zu bringen. Die Effektivität von Sicherheitsarchitekturen in service-orientierten Systemen hängt maßgeblich von der richtigen Nutzung und Integration von Security-APIs durch eine heterogene Gruppe von Softwareentwicklern ab, bei der nicht per se ein fundiertes Hintergrundwissen über komplexe digitale Sicherheitsmechanismen vorausgesetzt werden kann. Die Diskrepanz zwischen komplexen und in der Anwendung fehleranfälligen APIs und einem fehlenden Verständnis für die zugrundeliegenden Sicherheitskonzepte auf Seiten der Nutzer begünstigt in der Praxis unsichere Softwaresysteme. Aus diesem Grund ist die Gebrauchstauglichkeit von Security-APIs besonders relevant, damit Programmierer den benötigten Funktionsumfang effektiv, effizient und zufriedenstellend verwenden können. Abgeleitet von dieser Problemstellung, konzentriert sich das Dissertationsvorhaben auf die gebrauchstaugliche Ausgestaltung von Security-APIs und den Herausforderungen die sich aus den Methoden zur Evaluation der Usability in typischen Umgebungen der Softwareentwicklung ergeben.
Consolidating Principles and Patterns for Human-centred Usable Security Research and Development
(2018)
We present an evaluation of usable security principles and patterns to facilitate the transfer of existing knowledge to researchers and practitioners. Based on a literature review we extracted 23 common usable security principles and 47 usable security patterns and identified their interconnection. The results indicate that current research tends to focus on only a subset of important principles. The fact that some principles are not yet addressed by any design patterns suggests that further work on refining these patterns is needed. We developed an online repository, which stores the harmonized principles and patterns. The tool enables users to search for relevant patterns and explore them in an interactive and programmatic manner. We argue that both the insights presented in this paper and the repository will be highly valuable for students for getting a good overview, practitioners for implementing usable security and researchers for identifying areas of future research.
One of the main aims of current social robotic research is to improve the robots’ abilities to interact with humans. In order to achieve an interaction similar to that among humans, robots should be able to communicate in an intuitive and natural way and appropriately interpret human affects during social interactions. Similarly to how humans are able to recognize emotions in other humans, machines are capable of extracting information from the various ways humans convey emotions-including facial expression, speech, gesture or text-and using this information for improved human computer interaction. This can be described as Affective Computing, an interdisciplinary field that expands into otherwise unrelated fields like psychology and cognitive science and involves the research and development of systems that can recognize and interpret human affects. To leverage these emotional capabilities by embedding them in humanoid robots is the foundation of the concept Affective Robots, which has the objective of making robots capable of sensing the user’s current mood and personality traits and adapt their behavior in the most appropriate manner based on that. In this paper, the emotion recognition capabilities of the humanoid robot Pepper are experimentally explored, based on the facial expressions for the so-called basic emotions, as well as how it performs in contrast to other state-of-the-art approaches with both expression databases compiled in academic environments and real subjects showing posed expressions as well as spontaneous emotional reactions. The experiments’ results show that the detection accuracy amongst the evaluated approaches differs substantially. The introduced experiments offer a general structure and approach for conducting such experimental evaluations. The paper further suggests that the most meaningful results are obtained by conducting experiments with real subjects expressing the emotions as spontaneous reactions.
Cryptographic API misuse is responsible for a large number of software vulnerabilities. In many cases developers are overburdened by the complex set of programming choices and their security implications. Past studies have identified significant challenges when using cryptographic APIs that lack a certain set of usability features (e.g. easy-to-use documentation or meaningful warning and error messages) leading to an especially high likelihood of writing functionally correct but insecure code.
To support software developers in writing more secure code, this work investigates a novel approach aimed at these hard-to-use cryptographic APIs. In a controlled online experiment with 53 participants, we study the effectiveness of API-integrated security advice which informs about an API misuse and places secure programming hints as guidance close to the developer. This allows us to address insecure cryptographic choices including encryption algorithms, key sizes, modes of operation and hashing algorithms with helpful documentation in the guise of warnings. Whenever possible, the security advice proposes code changes to fix the responsible security issues. We find that our approach significantly improves code security. 73% of the participants who received the security advice fixed their insecure code.
We evaluate the opportunities and challenges of adopting API-integrated security advice and illustrate the potential to reduce the negative implications of cryptographic API misuse and help developers write more secure code.
Software development is a complex task. Merely focussing on functional requirements is not sufficient any more. Developers are responsible to take many non-functional requirements carefully into account. Security is amongst the most challenging, as getting it wrong will result in a large user-base being potentially at risk. A similar situation exists for administrators. Security defaults have been put into place here to encounter lacking security controls. As first attempts to establish security by default in software development are flourishing, the question on their usability for developers arises.
In this paper we study the effectiveness and efficiency of Content Security Policy (CSP) enforced as security default in a web framework. When deployed correctly, CSP is a valid protection mean in a defence-in-depth strategy against code injection attacks. In this paper we present a first qualitative laboratory study with 30 participants to discover how developers deal with CSP when deployed as security default. Our results emphasize that the deployment as security default has its benefits but requires careful consideration of a comprehensive information flow in order to improve and not weaken security. We provide first insights to inform research about aiding developers in the creation of secure web applications with usable security by default.
Further development on globally convergent algorithms for solution of stationary network problems is presented. The algorithms make use of global non-degeneracy of Jacobi matrix of the system, composed of Kirchhoff's flow conservation conditions and transport element equations. This property is achieved under certain monotonicity conditions on element equations and guarantees an existence of a unique solution of the problem as well as convergence to this solution from an arbitrary starting point. In application to gas transport networks, these algorithms are supported by a proper modeling of gas compressors, based on individually calibrated physical characteristics. This paper extends the modeling of compressors by hierarchical methods of topological reduction, combining the working diagrams for parallel and sequential connections of compressors. Estimations are also made for application of topological reduction methods beyond the compressor stations in generic network problems. Efficiency of the methods is tested by numerical experiments on realistic networks.