Events: Data- och informationsteknik events at Chalmers University of TechnologyFri, 14 Aug 2020 16:54:20 +0200 on Research with Marija Furdek<p>Zoom</p><p>​​</p><strong>​</strong><span style="background-color:initial"><strong>​​Welcome to Chalmers AI Research Center, CHAIR Spotlight on Research​. We are continuing this series of AI short talks with Assistant Professor at the Department of Electrical Engineering at Chalmers, Marija Furdek. On Friday August 21st she will handle the subject How can machine learning make optical communication networks more secure?</strong></span><br /><div><br /></div> <div><span style="background-color:initial">The first series of CHAIR Spotlight Research talks have the theme “New Chalmers researchers on the spot!”, meaning that researchers that came to Chalmers in the last three years are giving these talks. ​​​<br /></span><span style="background-color:initial"> ​</span></div> <div><span style="background-color:initial"><strong>Abstract: </strong>Optical networks are critical infrastructure supporting vital services and are vulnerable to different types of malicious attacks targeting service disruption at the optical layer. Due to the diverse physical-layer effects of different attack techniques and the limitations and sparse placement of optical performance monitoring devices, such attacks are difficult to detect and their signatures are unknown. However, coherent receivers with digital signal processing capabilities can provide elaborate optical monitoring data collected at the destination node of each connection, and machine learning techniques can give an unprecedented insight into the intricate relationships among the various signal parameters. </span></div> <div><span style="background-color:initial"><br /></span></div> <div><span style="background-color:initial">This talk presents a machine learning framework for detection and identification of physical-layer attacks, based on experimental attack traces from an operator field-deployed testbed subjected to in-band jamming, out-of-band jamming, and polarization modulation attacks with varying intensities. We evaluate different ML techniques in terms of their accuracy, robustness to missing features, and scalability in processing experimental data. We discuss the performance and suitability of supervised, unsupervised and semi-supervised learning techniques, as well as their embedding into the network service lifecycle. We also investigate challenges related to the integration of these techniques in real-life network management systems.</span></div>  ​<div><h2 class="chalmersElement-H2"><a href="" target="_blank" title="Register for Marija Furdek on 21 Aug">Register here &gt;​</a></h2> <h3 class="chalmersElement-H3">About the speaker</h3> <div> </div> <div><br /></div> <div> <img src="/SiteCollectionImages/Centrum/CHAIR/events/Marija-Furdek-Prekratic_170x170px.jpg" class="chalmersPosition-FloatLeft" alt="" style="margin:5px" /><span style="background-color:initial">D</span><span style="background-color:initial">r. Furdek is an assistant professor at the Department of Electrical Engineering at Chalmers. She received her Docent degree in Optical Networking from KTH Royal Institute of</span><span style="background-color:initial"> Technology in 2017. Her research focuses on the design of high-performance networks supporting next generation services, encompassing issues related to the optical network architecture and control. Her ambition is to contribute to the development of optimized, cognitive and autonomous communication networks with a focus on physical-layer security and resilience to a wide range of faults. As (co)PI, work group/package leader and researcher, Marija has participated in several European and Swedish research projects with a wide network of collaborators from the industry and academia. </span></div> <div><br />She is the leader of the Vetenskapsrådet project ‘Safeguarding optical communication networks from cyber-security attacks’. She has co-authored more than 100 scientific publications in international journals and conferences, 5 of which received best paper awards. Marija is an editor of the IEEE/OSA Journal of Optical Communications and Networking and the Photonic Network Communications journal, and a Guest Editor of the IEEE/OSA Journal of Lightwave Technologies. She is a General Chair of the IEEE Optical Network Design and Modeling (ONDM2021). She is a Senior Member of IEEE and OSA. <div><br /></div> ​</div></div>ño.aspx Gómez Londoño, Computer Science and Engineering<p>Online, link above</p><p>Choreographies and Cost Semantics for Reliable Communicating Systems</p><br /><div>Communicating systems have become ubiquitous in today's society. Unfortunately, the complexity of their interactions makes them particularly prone to failures such as deadlocked states caused by misbehaving components, or memory exhaustion due to a surge in message traffic (malicious or not). These vulnerabilities constitute a real risk to users, with consequences ranging from minor inconveniences to the possibility of loss of life and capital. This thesis presents two results that aim to increase the reliability of communicating systems. First, we implement a choreography language which by construction can only describe systems that are deadlock-free. Second, we develop a cost semantics to prove programs free of out-of-memory errors. Both of these results are formalized in the HOL4 theorem prover and integrated with the CakeML verified stack.</div> on Research with Karinne Ramirez-Amaro<p>Zoom (online)</p><p></p><b>​Welcome to Chalmers AI Research Center, CHAIR Spotlight on Research . In this series of AI short talks Karinne Ramirez-Amaro Assistant Professor at the department of </b><b><span></span>Electrical Engineering at Chalmers, will handle the subjetct: AI meets Robotics - Robots that Reason about Human Activities, via Zoom on Friday August 28.</b><div><strong></strong><b><br /></b> The first series of CHAIR Spotlight on Research talks have the theme “New Chalmers researchers on the spot!”, meaning that researchers that came to Chalmers in the last three years are giving these talks. ​​​</div> <div><br /></div> <div> <strong>Abstract: </strong>Autonomous robots are expected to learn new skills and to re-use past experiences in different situations as efficient, intuitive and reliable as possible. Robots need to adapt to different sources of information, for example, videos, robot sensors, virtual reality, etc. Then, to advance the research in the understanding of human activities, in robotics, the development of learning methods that adapt to different sensors are needed. In this talk, I will introduce a novel learning method that generates compact and general semantic models to infer human activities. This learning method allows robots to obtain and determine a higher-level understanding of a demonstrator’s behavior via semantic representations. First, the low-level information is extracted from the sensory data, then a meaningful semantic description, the high-level, is obtained by reasoning about the intended human behaviors. The introduced method has been assessed on different robots, e.g. the iCub, REEM-C, PR2, and TOMM, with different kinematic chains and dynamics. Furthermore, the robots use different perceptual modalities, under different constraints and in several scenarios ranging from making a sandwich to driving a car assessed on different domains (home-service and industrial scenarios). One important aspect of our approach is its scalability and adaptability toward new activities, which can be learned on-demand. Overall, the presented compact and flexible solutions are suitable to tackle complex and challenging problems for autonomous robots.</div> <div><br /></div> <h2 class="chalmersElement-H2">Registration for this event will open in August</h2> <div><br /> <div> <h3 class="chalmersElement-H3"><span>About the speaker<br /></span></h3> <img src="/SiteCollectionImages/Centrum/CHAIR/events/karinne_ramires_170px.jpg" class="chalmersPosition-FloatLeft" alt="" style="margin:5px 15px" />Dr. Karinne Ramirez Amaro is an Assistant professor at Chalmers University of Technology since September 2019. Previously, she was a post-doctoral researcher at the Chair for Cognitive Systems (ICS) at the Technical University of Munich (TUM). She completed her Ph.D. (summa cum laude) at the Department of Electrical and Computer Engineering at the Technical University of Munich (TUM), Germany in 2015. From October 2009 until Dec 2012, she was a member of the Intelligent Autonomous Systems (IAS) group headed by Prof. Michael Beetz. She received a Master degree in Computer Science (with honours) at the Center for Computing Research of the National Polytechnic Institute (CIC-IPN) in Mexico City, Mexico in 2007. Dr. Ramirez-Amaro received the Laura Bassi award granted by TUM and the Bavarian government to conduct a one-year research project in December 2015. </div> <div><br /></div> <div>For her doctoral thesis, she was awarded the price of excellent Doctoral degree for female engineering students, granted by the state of Bavaria, Germany in September 2015. In addition, she was granted a scholarship for a Ph. D. research by DAAD – CONACYT and she received the Google Anita Borg scholarship in 2011. She was involved in the EU FP7 project Factory-in-a-day and in the DFG-SFB project EASE. Her research interests include Artificial Intelligence, Semantic Representations, Assistive Robotics, Expert Systems, and Human Activity Recognition and Understanding.<br /> <h3 class="chalmersElement-H3">About CHAIR Spotlight on Research</h3> Chalmers AI Research Center, CHAIR Spotlight on Research is a series of AI short talks hosting researchers from Chalmers. The seminars are targeted towards experts from the Chair Consortium core partners as well as other Chalmers researchers. </div> <div><br /></div> <div> Our aim is to increase awareness of AI at Chalmers between Chalmers researchers and AI experts in industry. In the seminars, speakers present an overview of their current research and thoughts for new research, ideas, challenges – anything they believe to be of interest for other researchers. </div> <div><br /></div> <div>The seminar is taking place online and is scheduled to contain 30 minutes of presentation and 15 minutes of discussion. The seminars are open to all and are free of charge. CHAIR Spotlight Research talks are taking place on Fridays 13:00-13:45.</div> ​</div> Scheuner, Computer Science and Engineering<p>Online</p><p>​Towards Measuring and Understanding Performance in Infrastructure- and Function-as-a-Service Clouds</p><h3 class="chalmersElement-H3">Abstract </h3> <div><strong>Context. </strong>Cloud computing has become the de facto standard for deploying modern software systems, which makes its performance crucial to the efficient functioning of many applications. However, the unabated growth of established cloud services, such as Infrastructure-as-a-Service (IaaS), and the emergence of new services, such as Function-as-a-Service (FaaS), has led to an unprecedented diversity of cloud services with different performance characteristics. </div> <div><strong>Objective.</strong> The goal of this licentiate thesis is to measure and understand performance in IaaS and FaaS clouds. My PhD thesis will extend and leverage this understanding to propose solutions for building performance-optimized FaaS cloud applications. </div> <div><strong>Method.</strong> To achieve this goal, quantitative and qualitative research methods are used, including experimental research, artifact analysis, and literature review. Findings. The thesis proposes a cloud benchmarking methodology to estimate application performance in IaaS clouds, characterizes typical FaaS applications, identifies gaps in literature on FaaS performance evaluations, and examines the reproducibility of reported FaaS performance experiments. The evaluation of the benchmarking methodology yielded promising results for benchmark-based application performance estimation under selected conditions. Characterizing 89 FaaS applications revealed that they are most commonly used for short-running tasks with low data volume and bursty workloads. The review of 112 FaaS performance studies from academic and industrial sources found a strong focus on a single cloud platform using artificial micro-benchmarks and discovered that the majority of studies do not follow reproducibility principles on cloud experimentation. <br /></div> <div><strong>Future Work.</strong> Future work will propose a suite of application performance benchmarks for FaaS, which is instrumental for evaluating candidate solutions towards building performance-optimized FaaS applications.</div> <br /> Besker, Computer Science and Engineering<p>Online defence, available via link above</p><p>​Technical Debt: An empirical investigation of its harmfulness and on management strategies in industry</p><h3 class="chalmersElement-H3">​Abstract</h3> <div><strong>Background</strong>: In order to survive in today's fast-growing and ever fast-changing business environment, software companies need to continuously deliver customer value, both from a short- and long-term perspective. However, the consequences of potential long-term and far-reaching negative effects of shortcuts and quick fixes made during the software development lifecycle, described as Technical Debt (TD), can impede the software development process. </div> <div><strong>Objective</strong>: The overarching goal of this Ph.D. thesis is twofold. The first goal is to empirically study and understand in what way and to what extent, TD influences today’s software development work, specifically with the intention to provide more quantitative insight into the field. Second, to understand which different initiatives can reduce the negative effects of TD and also which factors are important to consider when implementing such initiatives. <br /></div> <div><strong>Method</strong>: To achieve the objectives, a combination of both quantitative and qualitative research methodologies are used, including interviews, surveys, a systematic literature review, a longitudinal study, analysis of documents, correlation analysis, and statistical tests. In seven of the eleven studies included in this Ph.D. thesis, a combination of multiple research methods are used to achieve high validity. </div> <div><strong>Results</strong>: We present results showing that software suffering from TD will cause various negative effects on both the software and the developing process. These negative effects are illustrated from a technical, financial, and a developer’s working situational perspective. These studies also identify several initiatives that can be undertaken in order to reduce the negative effects of TD. </div> <div><strong>Conclusion</strong>: The results show that software developers report that they waste 23% of their working time due to experiencing TD and that TD required them to perform additional time-consuming work activities. This study also shows that, compared to all types of TD, architectural TD has the greatest negative impact on daily software development work and that TD has negative effects on several different software quality attributes. Further, the results show that TD reduces developer morale. Moreover, the findings show that intentionally introducing TD in startup companies can allow the startups to cut development time, enabling faster feedback and increased revenue, preserve resources, and decrease risk and thereby contribute to beneficial effects. This study also identifies several initiatives that can be undertaken in order to reduce the negative effects of TD, such as the introduction of a tracking process where the TD items are introduced in an official backlog. The finding also indicates that there is an unfulfilled potential regarding how managers can influence the manner in which software practitioners address TD.</div> Abrahamsson, Computer Science and Engineering<p>Online</p><p>​Verified proof checking for higher-order logic</p><div>This thesis is about verified computer-aided checking of mathematical proofs. We build on tools for proof-producing program synthesis, and verified compilation, and a verified theorem proving kernel. Using these tools, we have produced a mechanized proof checker for higher-order logic that is verified to only accept valid proofs. To the best of our knowledge, this is the only proof checker for HOL that has been verified to this degree of rigor. <br /> </div> <div> Mathematical proofs exist to provide a high degree of confidence in the truth of statements. The level of confidence we place in a proof depends on its correctness. This correctness is usually established through proof checking, performed either by human or machine. One benefit of using a machine for this task is that the correctness of the machine itself can be proven. <br /></div> <div>The main contribution of this work is a verified mechanized proof checker for theorems in higher-order logic (HOL). The checker is implemented as functions in the logic of the HOL4 theorem prover, and it comes with a soundness result, which states that it will only accept proofs of true theorems of HOL. Using a technique for proof-producing code generation (which is extended as part of this thesis), we synthesize a CakeML program that is compiled using the CakeML compiler. The CakeML compiler is verified to preserve program semantics. As a consequence, we are able to obtain a soundness result about the machine code which implements the proof checker.</div> Maro, Computer Science and Engineering<p>Online defence</p><p>​Improving software traceability tools and processes</p> Schroeder, Computer Science and Engineering<p>Online defence.</p><p>​Understanding, Measuring and Evaluating Maintainability of Automotive Software</p> Poirot, Computer Science and Engineering<p>Online seminar, link above.</p><p>Coordination and Self-Adaptive Communication Primitives for Low-Power Wireless Networks</p><div><br /></div> <div>The Internet of Things (IoT) is a recent trend where objects are augmented with computing and communication capabilities, often via low-power wireless radios. The Internet of Things is an enabler for a connected and more sustainable modern society: smart grids are deployed to improve energy production and consumption, wireless monitoring systems allow smart factories to detect faults early and reduce waste, while connected vehicles coordinate on the road to ensure our safety and save fuel. Many recent IoT applications have stringent requirements for their wireless communication substrate: devices must cooperate and coordinate, must perform efficiently under varying and sometimes extreme environments, while strict deadlines must be met. Current distributed coordination algorithms have high overheads and are unfit to meet the requirements of today's wireless applications, while current wireless protocols are often best-effort and lack the guarantees provided by well-studied coordination solutions. Further, many communication primitives available today lack the ability to adapt to dynamic environments, and are often tuned during their design phase to reach a target performance, rather than be continuously updated at runtime to adapt to reality. </div> <div>In this thesis, we study the problem of efficient and low-latency consensus in the context of low-power wireless networks, where communication is unreliable and nodes can fail, and we investigate the design of a self-adaptive wireless stack, where the communication substrate is able to adapt to changes to its environment. We propose three new communication primitives: Wireless Paxos brings fault-tolerant consensus to low-power wireless networking, STARC is a middleware for safe vehicular coordination at intersections, while Dimmer builds on reinforcement learning to provide adaptivity to low-power wireless networks. We evaluate in-depth each primitive on testbed deployments and we provide an open-source implementation to enable their use and improvement by the community.</div> <br /> Giaimo, Computer Science and Engineering<p>Online, link above.</p><p>Bridging the Experimental Gap: Applying Continuous Experimentation to the Field of Cyber-Physical Systems, in the Example of the Automotive Domain</p><div><br /></div> <div>In the software world frequent updates and fast delivery of new features are needed by companies to bring value to customers and not lag behind competition. When in cyber-physical systems the software functionality dominates in importance the hardware capabilities, the same speed in creating new value is needed by the product owners to differentiate their products and attract customers. The automotive field is an example of a domain that will face this challenge as the industry races to achieve self-driving vehicles, which will necessarily be software-intensive highly complex cyber-physical systems.</div> A software engineering practice capable of accelerating and guiding the software production process using real-world data is Continuous Experimentation. This practice proved to be valuable in software-intensive web-based systems, allowing data-driven software evolution. It involves the use of experiments, which are instrumented versions of the software to be tested, deployed to the actual systems and executed in a limited way alongside the official software version. Valuable data on the future behavior of the prospective feature is collected in this way as it was fed the same real-world data it would encounter once approved and deployed. Additionally, in those cases where an experimental software version can be run as a replacement for the official version, relevant data regarding the system-user interaction can be gathered. In this thesis, the field of cyber-physical systems and the automotive practitioners' perspective on Continuous Experimentation are sampled employing a literature review and a series of case studies. A set of necessary architectural characteristics are defined and possible methods to overcome the issue of resource constraints in cyber-physical systems are proposed in two exploratory studies. Finally, a design study shows and analyses a prototype of a Continuous Experimentation cycle that was designed and executed in a project partnered by Revere, the Chalmers University of Technology's laboratory for vehicle research. Sjösten, Computer Science and Engineering<p>Online defence.</p><p>​Information Flow for Web Security and Privacy</p><div><br /></div> <div>The use of libraries is prevalent in modern web development. But how to ensure sensitive data is not being leaked through these libraries? This is the first challenge this thesis aims to solve. We propose the use of information-flow control by developing a principled approach to allow information-flow tracking in libraries, even if the libraries are written in a language not supporting information-flow control. The approach allows library functions to have unlabel and relabel models that explain how values are unlabeled and relabeled when marshaled between the labeled program and the unlabeled library. The approach handles primitive values and lists, records, higher-order functions, and references through the use of <em>lazy marshaling</em>. </div> <div>Web pages can combine benign properties of a user's browser to a fingerprint, which can identify the user. Fingerprinting can be intrusive and often happens without the user's consent. The second challenge this thesis aims to solve is to bridge the gap between the principled approach of handling libraries, to practical use in the information-flow aware JavaScript interpreter JSFlow. We extend JSFlow to handle libraries and be deployed in a browser, enabling information-flow tracking on web pages to detect fingerprinting. <br /></div> <div>Modern browsers allow for browser modifications through browser extensions. These extensions can be intrusive by, e.g., blocking content or modifying the DOM, and it can be in the interest of web pages to detect which extensions are installed in the browser. The third challenge this thesis aims to solve is finding which browser extensions are executing in a user's browser, and investigate how the installed browser extensions can be used to decrease the privacy of users. We do this by conducting several large-scale studies and show that due to added security by browser vendors, a web page may uniquely identify a user based on the installed browser extension alone. </div> <div>It is popular to use filter lists to block unwanted content such as ads and tracking scripts on web pages. These filter lists are usually crowd-sourced and mainly focus on English speaking regions. Non-English speaking regions should use a supplementary filter list, but smaller linguistic regions may not have an up to date filter list. The fourth challenge this thesis aims to solve is how to automatically generate supplementary filter lists for regions which currently do not have an up to date filter list.</div> <br /> Börjeson, Computer Science and Engineering<p>TBA</p><p>​Implementation of Energy-Efficient Carrier Phase Recovery Circuits for Optical Communication</p> Talks: Arthur Gretton<p>Zoom</p><p>Learning to generate realistic-looking images</p><p><br /><strong></strong></p> <strong>Critics for generative adversarial networks: results and conjectures<br /></strong> <p><strong>Abstract: </strong>Generative adversarial networks (GANs) use neural networks as generative models, creating realistic images that mimic real-life reference samples (for instance, images of faces, bedrooms, and more). These networks require an adaptive critic function while training, to teach the generator network how to improve its performance. To achieve this, the critic needs to measure how close generated samples are to true samples, and to provide a useful gradient signal the generator network.</p> <p>I will explore the design of critic functions for GANs, including f-divergences as used in the original GAN design, and integral probability metrics (such as the Maximum Mean Discrepancy) as used in later GANs. I will provide some observations and conjectures on critic design: in particular, a problem-specific critic seems to be helpful, and the critic needs to be deliberately weakened to ensure good GAN performance.</p> <p></p> <h2 class="chalmersElement-H2">About the speaker</h2> <p><img class="chalmersPosition-FloatRight" src="/SiteCollectionImages/Centrum/CHAIR/events/AI_Talks_ArthurGretton_180px_rev.jpg" alt="" style="margin:5px" /> </p> <p>Arthur Gretton is a Professor with the Gatsby Computational Neuroscience Unit, and director of the Centre for Computational Statistics and Machine Learning (CSML) at UCL. He received degrees in Physics and Systems Engineering from the Australian National University, and a PhD with Microsoft Research and the Signal Processing and Communications Laboratory at the University of Cambridge. He previously worked at the MPI for Biological Cybernetics, and at the Machine Learning Department, Carnegie Mellon University.</p> <p>Arthur's recent research interests in machine learning include the design and training of generative models, both implicit (e.g. GANs) and explicit (high/infinite dimensional exponential family models), nonparametric hypothesis testing, and kernel methods.</p> <p>He has been an associate editor at IEEE Transactions on Pattern Analysis and Machine Intelligence from 2009 to 2013, an Action Editor for JMLR since April 2013, an Area Chair for NeurIPS in 2008 and 2009, a Senior Area Chair for NeurIPS in 2018, an Area Chair for ICML in 2011 and 2012, a member of the COLT Program Committee in 2013, and a member of Royal Statistical Society Research Section Committee since January 2020. Arthur was program co-chair for AISTATS in 2016, tutorials co-chair for ICML 2018, workshops co-chair for ICML 2019, program co-chair for the Dali workshop in 2019, and co-organsier of the Machine Learning Summer School 2019 in London.</p>