DX'17: Editor's Preface

The International Workshop on Principles of Diagnosis (DX) is an annual event with a long tradition that started in 1989. It originated from the Artificial Intelligence community and has been focusing on theories, principles and computational techniques for diagnosis, monitoring, testing, reconfiguration & repair of complex systems, as well as applications of these techniques to real world problems.

It was our pleasure to organize the 28th DX that took place in Brescia, Italy, from September 26th to 29th. Of course there have been a lot of people involved in making a four day event happen. Two important parties to which we would like to express our gratitude are the program committee and the authors. For DX’17, we received 36 submissions (33 papers and 3 Ph.D. panel papers) from 11 countries and 4 continents. Each paper was thoroughly peer-reviewed by three reviewers. We wish to thank all the authors of submitted papers, and the program committee members for the time and efforts spent.

Of the Ph.D. panel submissions, all three were accepted for presentation at the Ph.D. panel chaired by Philippe Dague, where Gautam Biswas, Gregory Provan, and Belarmino Pulido completed the expert panel that lent their advice to the presenters. Two of the submissions are part of these proceedings. Out of the 33 regular submissions, 21 (63,6%) appear in these proceedings, with an additional five (15,2%) being presented at the workshop only. For the latter, please see our list below containing the authors as well as the title and a short summary for each presentation.

We owe a special debt of gratitude to the invited speakers, Anika Schumann from IBM Research Zurich, and Alexander Feldman from PARC, who contributed with their talks substantially to making DX’17 the interesting event that 34 people from 4 continents could experience, along with some local participants.

We would like to thank the team at FBK in Trento that took care of administration and registration issues, especially Annalisa Armani. Many thanks go also to Roxane Koitz, from TU Graz, and Roberto Zanotti, from the University of Brescia, for their local support throughout the workshop. We would like to express our gratitude also to the whole staff of our venue, the Centro Pastorale Paolo VI in Brescia, for their support and hospitality offered throughout the event, and the staff of EasyChair for their support with the reviewing process and publication of these proceedings.

DX'17 was under the patronage of the Comune di Brescia, which we gratefully acknowledge. We also would like to thank our sponsors: FBK – Fondazione Bruno Kessler, TU Graz – Institute for Software Technology, University of Brescia – Department of Information Engineering, Brescia Trasporti, and ATS – Advanced Technological Systems.

Last but not least, we would like to congratulate Philippe Dague. We are proud to confirm that we bestowed a well-deserved DX'17 lifetime achievements award to him for his contributions to the scientific field and his support of the DX community. Congratulations!

Please, let us conclude by underlying again that we are grateful for all for the support we received, and we hope that you will enjoy reading the proceedings of the 28th International Workshop on Principles of Diagnosis (DX'17)!

Your DX'17 chairs.

 


Additional presentations at DX'17 (in alphabetical order of the title):

Alexander Feldman and Johan de Kleer: Diagnosing Multiple Faults in Sequential Logic. Faults in physical or software systems can propagate in multiple ways: they can grow in time, show intermittent behavior, partially or fully self repair, or move from one component to another. Until now, all approaches from Artificial Intelligence (AI) for diagnosis of multiple faults had to assume some temporal pattern of behavior to prevent algorithmic combinational explosions. We propose a method that does not make any assumptions about the temporal or spatial evolution of multiple faults. This method works on systems where all components are driven by a single clock such as digital chips, micro-processors, and software systems. To overcome the combinational explosion of Model-Based Diagnosis (MBD) we use stochastic search. Even stochastic search is slow in diagnosing multiple failures in digital circuits. Therefore we use novel timebased batch update. We have evaluated the proposed algorithm on a benchmark of ISCAS-89 sequential circuits. The proposed approximation algorithm computes minimal-cardinality diagnoses with cardinality within 10% of the cardinality computed by state-of-the-art stochastic diagnostic search and only at a fraction of the time: between 2 and 200 times faster.

Roxane Koitz and Franz Wotawa: Extending the Modeling Framework for Abductive Diagnosis beyond Horn Clauses. Model-based diagnosis relies on a formalization of the system to be diagnosed, which has to characterize the underlying artifact accurately, provide information vital for the diagnostic task, and allow computing solutions efficiently. A common constraint of diagnosis engines is to restrict the model to comprise Horn clauses. There are, however, practical applications where this limitation is problematic as the failure behavior of the system would require a more expressive representation. In this paper, we describe a modeling method for abductive diagnosis based on an extension of Horn logic, which allows expressing conjunctions and disjunctions of effects. We present a mapping from this representation to a Horn theory since such a model can be applied to a wide range of diagnosis algorithms. Diagnosis based on the converted models provides intuitive explanations similar to human reasoning, which are beneficial in the context of practical fault identification. In addition, we present initial results comparing diagnosis on two well-known abductive inference engines in regard to the Horn transformation.

Belarmino Pulido, Raymond Sterling, Jorge Hernández Benito, Carlos Alonso-González and Marcus Keane: HVAC diagnosis using Hybrid Possible Conflicts. Heating, Ventilation and Air Conditioning systems demand continuous operation during long time intervals. These systems are a clear example of hybrid systems because they work under several operation modes depending on the environmental conditions where they are deployed, but also depending in the intended set points for both humidity and temperature. Model-based diagnosis of hybrid systems can be accomplished using different modelling techniques and can be approached using different paradigms. In this work we propose to perform Consistency-based diagnosis with Hybrid Possible Conflicts (HPCs), that allows to track continuous behaviour and that is able to cope with both discrete and parametric faults in the systems in an unified framework. To achieve this goal we have analysed existing detailed numerical Modelica models to derive HPC simulation models whose associated residuals can be used to perform fault detection and isolation; moreover we have analysed the qualitative effects of both parametric and discrete faults in the HPCs to further discriminate among potential fault candidates. This approach has been tested on an Air Handling Unit, AHU-9, whose operation modes has been analysed and several diagnosis scenarios have been tested.
 
Youssef Youssef, Chaitanya Hebbal, Ahmad Drak, Anastassia Küstenmacher and Paul Plöger: Model Based Remote Diagnosis of Motion Faults on an Omnidirectional Robot via Structural Analysis. The ability of diagnosing faults is an integral part in enhancing the reliability of mobile platforms. However, striving to attain high reliability of the platform highly increases the complexity of the system. Therefore, the use of a non-intrusive robotic black box reduces the complexity of the system and, in return, saves valuable computing resources. In this study, remote fault diagnosis is applied to an omnidirectional robotic platform with Mecanum wheels to diagnose motion faults. A model-based approach is used for fault detection, and structural analysis is used for fault isolation and identification. The dynamic model of an omnidirectional platform is derived, validated and then tested for fault detection, and the structural analysis relations are obtained and discussed.

Sofia Reis and Rui Abreu: Using GitHub to Create a Dataset of Natural Occurring Vulnerabilities. Currently, to satisfy the potential high number of system requirements, complex software is crafted which turns its development cost-intensive and more susceptible to security vulnerabilities. According to IBM's X-Force Threat Intelligence 2017 Report, the number of vulnerabilities per year has been significantly increasing over the past years. In software security testing, performing empirical studies is challenging due to the lack of widely accepted and easy-to-use databases of real vulnerabilities as well as the fact that it requires both human effort and CPU time. Consequently, researchers tend to use databases of hand-seeded vulnerabilities, which may differ inadvertently from real vulnerabilities and thus might lead to misleading assessments of the capabilities of the tools. Although there are databases targeting security vulnerabilities test cases, only one database contains real vulnerabilities, the other ones are a mix of real and artificial or even only artificial samples. This paper explains our efforts to create a vulnerability database, Secbench, by mining 238 repositories from GitHub. GitHub is particularly interesting since it hosts millions of open-source projects carrying a considerable number of security vulnerabilities. More than 1M of commits were mined for 16 different patterns which yielded 602 security vulnerabilities. The study described in this paper provides a methodology to mining security vulnerabilities from open-source software. Our methodology has proven itself as being valuable since we were able to collect a considerable number of security vulnerabilities from a small group of repositories. However, there is still much work to do in order to improve not only the mining process but also the vulnerabilities diagnosis. All the information related to Secbench is available at/through https://tqrg.github.io/secbench/.