Search

Data in business processes is becoming more and more important. Current standards for process-modeling languages like BPMN 2.0 which include the data flow reflect this. Ensuring the correctness of the data flow in processes is challenging. Model checking, i. e., verifying properties of process models, is a well-known technique to this end. An important part of model checking is the construction of the state space of the model. State-space explosion however typically is in the way of an effective verification. We study how to overcome this problem in our context by means of reduction. More specifically, we propose a reduction on the level of the process model. To our knowledge, this is new for the data-flow analysis of processes. To accomplish this, we specify regions relevant for the verification of properties describing the data flow. Our evaluation shows that our approach works well on real process models.

Information hiding ensures privacy by transforming personalized data so that certain sensitive information cannot be inferred any more. One state-of-the-art information-hiding approach is the Pufferfish framework. It lets the users specify their privacy requirements as so-called discriminative pairs of secrets, and it perturbs data so that an adversary does not learn about the probability distribution of such pairs. However, deploying the framework on complex data such as time series requires application specific work. This includes a general definition of the representation of secrets in the data. Another issue is that the tradeoff between Pufferfish privacy and utility of the data is largely unexplored in quantitative terms. In this study, we quantify this tradeoff for smart meter data. Such data contains fine-grained time series of power-consumption data from private households. Disseminating such data in an uncontrolled way puts privacy at risk. We investigate how time series of energy consumption data must be transformed to facilitate specifying secrets that Pufferfish can use. We ensure the generality of our study by looking at different information-extraction approaches, such as re-identification and non-intrusive-appliance-load monitoring, in combination with a comprehensive set of secrets. Additionally, we provide quantitative utility results for a real-world application, the so-called local energy market.

Business process design for real-world applications often requires modelling languages of a certain complexity, interweaving several process perspectives, such as the data and the organizational one, and taking contextual process information into account. Ensuring that such languages are indeed usable is not trivial. This paper describes a usability study for a modelling language with those characteristics. More specifically, the so-called BTG! language features the security concept Break the Glass (BTG) that allows to specify exceptional access to data. We investigate usability characteristics of BTG! such as learnability, efficiency, errors, and satisfaction. Our study has helped to increase the usability of the language significantly. To illustrate, we have replaced the language terms with more intuitive ones and have extended the language with some frequently desired features. Our experiences can be applied to process modelling languages that are similarly comprehensive.

Information hiding ensures privacy by transforming personalized data so that certain sensitive information cannot be inferred any more. One state-of-the-art information-hiding approach is the Pufferfish framework. It lets the users specify their privacy requirements as so-called discriminative pairs of secrets, and it perturbs data so that an adversary does not learn about the probability distribution of such pairs. However, deploying the framework on complex data such as time series requires application specific work. This includes a general definition of the representation of secrets in the data. Another issue is that the tradeoff between Pufferfish privacy and utility of the data is largely unexplored in quantitative terms. In this study, we quantify this tradeoff for smart meter data. Such data contains fine-grained time series of power-consumption data from private households. Disseminating such data in an uncontrolled way puts privacy at risk. We investigate how time series of energy consumption data must be transformed to facilitate specifying secrets that Pufferfish can use. We ensure the generality of our study by looking at different information-extraction approaches, such as re-identification and non-intrusive-appliance-load monitoring, in combination with a comprehensive set of secrets. Additionally, we provide quantitative utility results for a real-world application, the so-called local energy market.

Process models often are generic, i. e., describe similar cases or contexts. For instance, a process model for commissioning can cover both vehicles with an automatic and with a manual transmission, by executing alternative tasks. A generic process model is not optimal compared to one tailored to a specific context. Given a declarative specification of the constraints and a specific context, we study how to automatically generate a good process model and propose a novel approach. We focus on the restricted case that there are not any repetitions of a task, as is the case in commissioning and elsewhere, e. g., manufacturing. Our approach uses a probabilistic search to find a good process model according to quality criteria. It can handle complex real-world specifications containing several hundred constraints and more than one hundred tasks. The process models generated with our scheme are superior (nearly twice as fast) to ones designed by professional modelers by hand.

uploaded by Plazi

(Semi-)automated bug localisation is an important issue in software engineering. Recent techniques based on call graphs and graph mining can locate bugs in relatively small programs, but do not scale for real-world applications. In this paper we describe a bug-localisation approach based on graph mining that has this property, at least according to preliminary experiments. Our main contribution is the definition and analysis of class-level call graphs, with encouraging results.

Data Mining Techniken wie Klassifikation, Regression und Clusteranalyse finden heutzutage eine weite Verbreitung. Entsprechende relationale Daten liegen in vielen Anwendungsdomänen vor, und effiziente Data Mining Algorithmen sind in kommerzielle Werkzeuge sowie in Datenbank Management Systeme integriert. In den letzten Jahren wurden aber auch verschiedene strukturelle Data Mining Techniken entwickelt, die z.~B.\ mit Graph-basierten Daten arbeiten. Solche Techniken erschließen neue Anwendungsgebiete, bieten aber auch das Potential, bisherige Techniken zu ergänzen. Oft können durch Kombination bisherige Ergebnisse verbessert werden. In diesem Beitrag präsentieren wir Arbeiten aus dem Bereich der Vorhersage von Kundenverhalten und der Fehlersuche in Software, in denen strukturelle und relationale Data Mining Techniken erfolgreich kombiniert wurden. Schließlich geben wir einen Ausblick auf weitere Anwendungsgebiete und zukünftige Herausforderungen.

This work addresses the problem of discovering bugs in software development. We investigate the utilisation of call graphs of program executions and graph mining algorithms to approach this problem. We propose a novel reduction technique for call graphs which introduces edge weights. Then, we present an analysis technique for such weighted call graphs based on graph mining and on traditional feature selection. Our new approach finds bugs which could not be detected so far. With regard to bugs which can already be localised, our technique also doubles the precision of finding them.

CS technical report, 340