Dumb Web, Smart Web, Knowledgable Web

Abstract

In February 2011, Watson (IBM super computer) managed to beat two past grand champions on the TV quiz show Jeopardy!. Watson was able to answer questions that require intelligence when done by humans. This marked the first machine to pass the Turing test and started a new era of computing called cognitive computing where computers (modeled after the human brain) learn and interact naturally with people to extend what either humans or machine could do on their own. Google, Amazon, Facebook, etc., using big data technologies, smart Machine Learning and AI algorithms were able to tap into our intensions by predicting what we click, buy, like, dislike, shop, surf, etc.,. This marked the era of smart Web or predictive intelligence where Turing test is reversed and machines now try to figure out who we are, our information needs, our behavioral patterns, the activities we are engaged in and our goals. Together with the torrents of data we leave behind us every time we communicate with the digital eco-system, a new era of human-machine cooperation is starting that gives us millions of potential insights into user experience, marketing, personal tastes, and human behavior. In this talk, we are going to illustrate through motivating cases and research directions the main characteristics of this era and how it can transform the way we interact with the Web to ultimately improve the quality of our lives and gain valuable insights into our affective, mental and physical states.

Abstract

Social Engineering has long been a very effective means of attacking information systems. The term knowledge worker has been coined by Peter Drucker more than 50 years ago and still describes very well the basic characteristics of many employees. Today, with current hypes such as BYOD (bring your own device) and public cloud services, young professionals expect to use the same technology both in their private life and while working. In global companies teams are no longer geographically co-located but staffed globally just-in-time. The decrease in personal interaction combined with the plethora of tools used (E-Mail, IM, Skype, Dropbox, Linked-In, Lync, etc.) create new opportunities for attackers. As recent attacks on companies such as the New York Times, RSA or Apple have shown, targeted spear-phishing attacks are an effective evolution of social engineering attacks. When combined with spear phishing to distribute zero-day-exploits they become a dangerous weapon, often used by advanced persistent threats. In this talk we will explore some attack vectors and possible steps to mitigate the risk.

Domain Specific Languages: From Craft to Engineering

Abstract

The engineering of systems involves many different stakeholders, each with their own domain of expertise. Hence more and more organizations are developing an ever growing number of Domain Specific Languages (DSLs) to allow domain experts to express solutions directly in terms of relevant domain concepts. This new trend raises new challenges about designing not just one DSL but many of them, evolving a set of DSLs and coordinating the use of multiple DSLs. In this talk we explore various dimensions of these challenges, and outline a possible research roadmap for addressing them. We detail one of these challenges, which is the safe reuse of model transformations. Indeed both DSL definition and tooling (eg. checkers, document or code generators, model transformations) require significant development efforts, for a limited audience (by definition), because the current state of the art of Model Driven Engineering still makes it hard to reuse and evolve these definitions and tooling across several DSLs, even when these DSLs are conceptually very close to one other. We outline a new extension to the Kermeta language that leverages Family Polymorphism to allow model polymorphism, inheritance among DSLs, as well as evolution and interoperability of DSLs.

Resilient Optical Network Virtualization

Abstract

Cloud computing services are emerging as an essential component of the industry ICT infrastructure and, consequently, one of the fastest growing business opportunities for Internet infrastructure and service providers. Many enterprises are moving their services towards cloud infrastructures. In this rapidly growing market, datacenter (DC) scalability is becoming a major technical challenge for service providers as well as its performance optimization, with a key focus on the network technologies and their control. In fact, service providers have to cope with cloud services delivered by more and more geographically distributed DCs, ever increasing requests by users and DC providers for very high throughputs and low latencies, resource dynamicity and elasticity (i.e. flexible storage and computing on demand) and seamless resource/service migration. The future Internet architecture needs to offer:

• An efficient integration between the high-performance and high-bandwidth optical network infrastructure of operators and services/resources provided by DCs and server farms.
• End-to-end cloud service provisioning eco-system that automatically and efficiently bundles DC infrastructure services (i.e. computing and storage) with the required operator optical network connectivity services.
• A highly resilient virtual infrastructure as a major disadvantage of a virtual architecture is that a single hardware fault or a defect in the virtualization infrastructure (VI) software can lead to the failure of multiple virtual nodes by, for example, preventing VMs from being scheduled or preventing VMs from accessing I/O devices. Hence, a key to enabling the deployment of virtual clusters is to enhance the resiliency of the VI to faults in the hardware or virtualization software.

Last-Mile Delivery in Smart Cities

Abstract

Logistics and transportation are two very aspects in smart city development. This talk is in 2 parts. I will discuss the challenges of urban logistics, more specifically in the context of last-mile delivery of freight into mega-cities like Singapore. I will first present several case studies in Europe and Asia on the operation of Urban Consolidation Centers (UCCs) which enable collaboration among shippers, carriers, and retailers to consolidate deliveries. I will then discuss my research on market mechanisms and system that provide the necessary technology and incentives for multiple stakeholders to derive win-win benefits from operating within UCCs. In the second part, I will discuss last-mile delivery of passengers from a transport hub (such as a major train/bus station) to their homes. I will discuss viable approaches from different urban cities in Asia and Europe. Then, I will discuss my research on last mile transportation. More specifically, I will present 2-level planning framework for making real-time clustering and routing decisions. The first upper/strategic level selects, from the set of current requests, the ones to be scheduled in a given period, which can be modelled and solved as Markov decision process (MDP). The second lower/operational level clusters and then routes the selected requests using a standard vehicle routing heuristic.

Exploring Linked Statistical Data Using Linked Widgets

Abstract

The Open Data movement has gained momentum among governments, in the business world, and in the public sector over the last few years. The result of this movement is a growing number of open and accessible datasets that have established a solid basis for offering enhanced services and improved experiences to citizens and business. The statistical data, which embodies a big portion of Open Data, comprises a wide range of domains including finance, demographics, transportation, employment, etc. Statistical data plays an increasingly important role in public policy formation and as a facilitator for informed decision-making in the private sector. In this context, Linked Statistical Data is an evolving domain that combines the power of Linked Data (a set of best practices for publishing and connecting structured data on the Web) with open statistical data in order to empower linkage of relevant data and concepts from different resources. In this talk, the Linked Statistical Data limitations and challenges for knowledge workers will be explored and then the innovative concept of Linked Widgets will be presented which is aiming to bridge the knowledge gap between statistical data publishers and knowledge workers who consume the statistical data.

Mending Fences under Weak memory Models

Abstract

This talk is devoted to the automatic verification and correction of concurrent programs running under weak memory models. These models capture the reordering of program (read and write) operations done by modern multi-processor architectures and compilers for performance reasons. Thus, the programs may exhibit behaviours that deviate substantially from their behaviours under the usual Sequentially Consistent (SC) (i.e., interleaving) semantics. The standard way to eliminate the undesired behaviours is to insert memory fences (or barriers) instructions that typically prevent reordering of some instructions issued before and after the fence. An important problem in concurrent programming is to find sets of fences that ensure the program correctness without compromising efficiency. This is an important and crucial problem for the correctness of concurrency libraries and other performance-critical system services employing lock-free synchronization, as well as for the correctness of compiler backends that generate code targeted to run on such architectures. In this talk, we consider the most common model resulting from the "write to read" relaxation, which corresponds to the TSO (Total Store Ordering) model. This relaxation is used in most hardware architectures available today. We define abstract operational models for this weak memory model based on state machines with (potentially unbounded) FIFO buffers. Then, we investigate the decidability and the complexity of checking safety properties for programs under TSO memory models. Finally, we present an algorithm that produces an optimal set of fences needed to ensure the program correctness with respect to a safety property. The algorithm is counter-example guided, using counter-examples that are produced by checking safety properties of programs running under TSO memory models.