SaS Seminars 2006
Software and Systems Research Seminar Series
Modeling for Diagnostics
Date: November 23 Place: Alan Turing Time: 15:15
Abstract: It's an all to familiar scene from a very famous movie: Apollo 13. The movie shows the cockpit of the Apollo 13 where astronaut Jim Lovell (played by Tom Hanks) reports back to earth: "Houston, we have a problem". The plot of the movie is based on the third American manned lunar landing mission, part of the Apollo program. Two days after the launch, a malfunction in the spacecraft caused an explosion that made one of the spacecrafts service modules to loose oxygen and electrical power. Astronauts John Swigert, Jr., James Lovell and Fred Haise Jr., who made up the crew of the US's Apollo 13 moon flight, used this phrase to report a life threatening technical problem:
Swigert(LMP): Okay, Houston
Lovell (CDR): I believe we've had a problem here.
Capcom (CC): This is Houston. Say again, please.
CDR: Houston, we've had a problem. We've had a main B bus undervolt.
CC: Roger. Main B undervolt
LMP: Okay, Right now, Houston, the voltage is - is looking good. And we had a pretty large bang associated with the CAUTION AND WARNING there. And as I recall, MAIN B was the one that had had an amp spike on it once before.
With increased system complexity in recent years almost every system under deployment is exposed to component failures or is under the risk of suffering a major breakdown under its life time, as the one suffered by Apollo 13. Maintenance and repair is an ever increasing part of the total cost of a final product. Diagnosis techniques have been adopted by the after market departments of industrial systems product developers as a fast and accurate way of finding the root causes of failures.
The seminar examines issues related to the development of integrated systems for system-level diagnosis. I will start by presenting the field technician's view of the problem and how they need is addressed by current software systems. Then, we will explore the fundamental issues and challenges in system diagnosis and prognostics from the computer scientist's point of view and discuss how the detection, localization and isolation of faults can be achieved by using statistical methods, decision theory, AI methods or constraint propagation techniques. In recent years, model-based diagnostics reasoning systems have provided a major advance in fault isolation and reduction of repair time contributing to the reduction of maintenance cost. Therefore, model based diagnosis techniques will receive a special attention during this seminar and their applicability will be illustrated on several real life industrial examples.
Challenges for spoken dialogue technology
Date: October 26 Place: Alan Turing Time: 15:15
Dr. Johan Boye , Telia Research AB, Farsta
Speech interfaces to commercial services are becoming more and more common. Many people today have experience from speaking with machines when booking tickets or retrieving information over the telephone. This development is largely due to advances in automatic speech recognition, i.e. technology for identifying the words of an utterance. However, the commercial services of today are still rather limited in their dialogue capabilities. A typical dialogue consists of a number of questions from the system, posed in a fixed order, to which the user can reply only in a very restricted way.
In order to exploit the true potential of natural language, it is necessary to develop spoken dialogue technology that allow more flexible system behavior. It is essential that users can express themselves more freely, and that dialogue systems understand the meaning of the user's utterances when uttered as part of an connected dialogue.
In this talk, I will discuss some of the current challenges in spoken language understanding and spoken dialogue management, and present work carried out at TeliaSonera to meet them. The proposed approaches will be discussed in the light of two research prototypes, a database retrieval application and a 3D computer game, as well as a recently deployed commercial application for automated customer care.
Johan Boye got his Ph.D. in logic programming at IDA in 1996, and is currently working at the research division at TeliaSonera in Stockholm. His main activities are research and application development in the field of spoken dialogue processing, with the aim of producing computer systems that can allow people to access various services using spoken natural language.
Toward Building Secure Systems
Date: October 5 Place: Alan Turing Time: 15:15
Prof. Dr. Christoph Schuba , LiU, IDA
Society is rapidly moving toward a future that's increasingly dependent on technology, especially hardware and software systems. There are more and more computers, cell phones, PDAs, and other devices connected to one another over a digital heartbeat, providing user, business, and infrastructure services. These distributed services only function because of an intricate interplay of systems and software most users are not even aware exist.
Security is one important quality that all these systems should possess. But what is meant by the term "security" in these different contexts and how can it be achieved? The term security is so hopelessly overloaded with meaning that its use captures a vast set of properties and problems we all care about today, in high technology and in society as a whole.
This presentation explains what the term security means in the context of computer systems. We then explore a number of approaches that contribute to varying degrees to the quest for the holy grail, system security. Along the way we will be pointing out open research problems and ongoing work.
Christoph Schuba studied mathematics and management information systems at the Universität Heidelberg and the Universität Mannheim in Germany. As a Fulbright scholar, he earned his M.S. and Ph.D. degrees in Computer Science from Purdue University in 1993 and 1997, performing most of his dissertation research in the Computer Science Laboratory at the Xerox Palo Alto Research Center (PARC). Christoph has taught undergraduate and graduate courses in computer and network security, cryptography, operating systems, and distributed systems at the Universtität Heidelberg, Germany, at the International University in Bruchsal, Germany, and at San Jose State University, USA. Since 1997 he has been working in Sun Labs and the Security Technology Office at Sun Microsystems, Inc. In January 2006, Christoph joined Linköpings Universitet in Sweden as Professor in Information Security.
Superinstructions and Replication in the Cacao JVM interpreter
Date: June 8 Place: Alan Turing Time: 10:15
Prof. Anton M. Ertl , TU Vienna, Austria
Abstract: Dynamic superinstructions and replication can provide large speedups over plain interpretation. In a JVM implementation we have to overcome two problems to realize the full potential of these optimizations: the conflict between superinstructions and the quickening optimization; and the non-relocatability of JVM instructions that can throw exceptions. In this paper, we present solutions for these problems. We also present empirical results: We see speedups of up to a factor of 4 from superinstructions with all these problems solved. The contribution of making potentially throwing JVM instructions relocatable is up to a factor of 2. Replication has a small, but usually positive effect on performance.
Software Analysis in Theory and Practice
Date: April 27, Place: Alan Turing, Time: 15:15
Professor Welf Löwe , Växjö University,
Abstract: Software maintenance includes tasks like finding and removing design flaws and adding new functionality to a software system. These maintenance tasks are expensive today. Cost estimations of resources and time range from 50% to 80% of the total costs of ownership of a software system. In order to maintain a system, it needs to be comprehended and the effort for gaining comprehension even dominates the total maintenance effort. Estimations range from 40% up to 90%.
Design documents and other kinds of documentation providing an abstract and comprehensible view on a system are often outdated or not available at all. In many cases, the code of the system is the only trustworthy source of information. However, due to the size of many systems, gaining comprehension from code needs to be supported by tools. The task of these tools is to extract information, analyze and focus it, and, finally, visualize it for the person who tries to understand the software system.
Depending on the maintenance goal, the abstraction level, and the humans involved, different views of a software system are appropriate. This implies that different visualizations need to be designed, different analyses to be defined, and different kinds of information to be extracted.
In this talk, we sketch some example analyses: Points-To Analysis is a static source code analysis computing possible dynamic call-, access-, and reference-relations between runtime entities like objects. Design Pattern Detection aims at identifying design and other code patterns in source code or other system specifications like UML diagrams. Architecture Recovery is the task of automatically analyzing the architecture of a system, i.e., detecting its components and connectors. Metric Analysis pushes the level of abstraction even further: they aim at numerically accessing certain properties and qualities of a system. Moreover, we present synergies between these analyses and give some practical applications.
Finally, we introduce a framework, the VizzAnalyzer, for integrating different kinds of analyses and visualizations in a simple and straightforward way.
Testing Stream Cipher Generators and Pseudorandom Number Generators by State Space Exploration
Date: March 23, Place: Alan Turing, Time: 15:15
Professor Joerg Keller, Fern-Universität in Hagen, Germany
Abstract: Both generators for stream ciphers (such as A5/1) or pseudorandom number generators are typically initialized into a certain state by a seed value, and then obtain a sequence of states they are in. In each state they output one or more bits. The state space can thus be modeled as a directed graph where each node has exactly one outgoing edge to the follow-up state. Each weakly connected component of such a graph consists of one cycle and a number of trees directed to the root, with the root being a node on the cycle. Often one is interested in the lengths of the cycles (period lengths), as they frequently cannot be derived analytically. Also, in crypto applications, one is interested in the graph structure as a whole. If the generator shall randomize, then its graph structure should not deviate too much from the expected values for a randomly chosen graph with outdegree exactly 1. We present a parallel algorithm to explore such a graph without constructing it in memory, which is unfeasable because of its size. We report on the experimental runtimes achieved, and we report on the graph structures revealed for some generators.
Page responsible: Christoph Kessler
Last updated: 2012-08-17