The performance and reliability we demand from switched networks, such as data center networks, backbone WAN, or even customer LAN, has grown by orders of magnitude. Such demands have been supported through advances in robust and high-performance switch hardware, transmission and media technologies. However, no matter how robust, network components still eventually fail. At the time of failure, network resilience and fast recovery depend on having built-in redundancy in the network, and fast, fault- tolerant Layer 2 protocols. A closer look reveals that there is a strong unmet need for such protocols. The existing protocols such as RSTP are simply not fast enough for Data Centers etc. (In response to this challenge, such networks give up L2 solutions, and build highly-resilient infrastructure with fault recovery etc. as L3 SDN solutions, or introduce L3 technology into L2, as with TRILL.) In this paper, we introduce a fast and robust loop-avoidance protocol for pure L2 networks, named the Meshed Tree Protocol (MTP); it provides resilience and fast recovery for switched networks.

Congestion in urban areas carries inherent costs in terms of fuel consumption, pollution, delays. Although ITS systems have devised solutions such as GLOSA, they require sophisticated infrastructure and thus installation and maintenance investments. In this paper, we propose V^3TL, a V2V-based distributed solution to manage unregulated intersections that aims primarily at minimizing the time needed to clear the intersection, while reducing the number of stop-and-go maneuvers that are known to increase emissions. The proposed solution operates a cyclic scheduling of vehicles as they approach the intersection, letting them coordinate among each other through V2V communication. We simulated the solution in an unregulated four-way intersection, showing that it achieves its objectives.

Remote key-less entry systems (KES) are one of the most important and desirable features of modern vehicles. However, due to platform dependencies and/or design flaws, majority of the KES used even in recent high end cars are vulnerable to signal relay attacks. In this work, we propose Proxicar - an architecture for secure Digital Key Solution that enables user to use an authorised digital key stored in his/her personal device for remote vehicle access. We study the feasibility of employing physical layer features i.e., Received Signal Strength Indicator (RSSI) of Low Power Bluetooth (BLE) technology for (i) proximity estimation, (ii) user activity detection, and (iii) second factor authentication between the vehicle and user device. We highlight the major challenges involved in using BLE channel features for security, and present a method to overcome those issues. We implement our proposed method on off-the-shelf android-based devices and experimentally evaluate the performance. Our promising results show that BLE physical layer features can be effectively employed to design secure KES.

Cyberphysical Systems (CPS) are transforming the way we interact with the physical world around us. However, centralised approaches for CPS systems are not capable of addressing the unique challenges of CPS due to the complexity, constraints, and dynamic nature of the interactions. To realize the true potential of CPS, a decentralized approach that takes into account these unique features is required. Recently, blockchain-based solutions have been proposed to address CPS challenges. Yet, applying blockchain for diverse CPS domains is not straightforward and has its own challenges. In this paper, we share our experiences in applying blockchain technology for CPS with an aim to provide insights and highlight the challenges and future opportunities.

Telecommunication networks are designed for high reliability; however, when they do fail, it is difficult to detect problems in a timely manner as the networks are characteristically complex and the problems may be subtle rather than catastrophic. Within the overall objective of network anomaly detection, the focus of this paper is on: dimensionality reduction through feature selection to determine the minimum set of features that provide the most information regarding the network state; and the application of the multivariate unsupervised learning technique, Principal Component Analysis (PCA), to detect anomalies with low detection latency. For anomaly detection, we apply PCA on the normal data, and find the subspace where the variation of the normal data is small. By characterizing the variation of the normal data in the subspace, we derive a boundary of the normal data, and develop an anomaly detection model based on it. The algorithms are developed and tested with Per Call Measurement Data records from a 4G-LTE network. The impact of our work is the proactive detection of network anomalies in mobile networks, thereby improving network reliability and availability.

Recently, it has been widely accepted by the research community that interactions between humans and cyber-physical infrastructures have played a significant role in determining the performance of the latter. The existing paradigm for designing cyber-physical systems for optimal performance focuses on developing models based on historical data. The impacts of context factors driving human system interaction are challenging and are difficult to capture and replicate in existing design models. As a result, many existing models do not or only partially address those context factors of a new design owing to the lack of capabilities to capture the context factors. This limitation in many existing models often causes performance gaps between predicted and measured results. We envision a new design environment, a cyber-physical human system (CPHS) where decision-making processes for physical infrastructures under design are intelligently connected to distributed resources over cyberinfrastructure such as experiments on design features and empirical evidence from operations of existing instances. The framework combines existing design models with context-aware design-specific data involving human-infrastructure interactions in new designs, using a machine learning approach to create augmented design models with improved predictive powers.

At Microsoft, we are developing new Search & AI technologies for various vertical markets, education being one of them. We look to reimagine input modalities through speech and camera to make search easily accessible on mobile devices without the need to input a query in a search box. Image as an input is particularly challenging, and we develop new ways to understand user intent and provide appropriate knowledge and actions geared towards learning and understanding concepts. Through these innovations, we have an education focused product that is easily accessible to students on their mobile devices.

Recognizing human activity and behavior is a well-known research problem. State-of-the-art machine learning approaches, a variety of smart home sensor systems, and Internet-of-Things (IoT) devices are in place to help aid activity discovery, recognition, and inference processes. However, the current approaches work with single user, a predefined set of devices and in a specific smart home environment. Scaling, adapting, and crossing the boundaries beyond those presumed settings in presence of limited or unlabeled datasets is a challenging research avenue to investigate. In this talk, I will first articulate the underpinning challenges across heterogenous IoT devices, different users, and environments. I will discuss our preliminary work on addressing those challenges using deep learning-based domain adaptation and transfer learning-based techniques, and present experimental results using real datasets collected from different IoT devices such as smartphones, smartwatches, and diverse users. I will also briefly touch upon the data analytics techniques we have been employing in some other research projects related to smart environments such as functional and cognitive health assessment of older adults, mobile gunshot detection, automated dance assessment, virtual energy audits, flash flood detection, and sports analytics.

This talk will first present a brief overview of some key new features of the about-to-be-finalized (late 2019) latest WiFi (802.11ax or WiFi6) standard, that aims to improve operational efficiency in dense deployment scenarios. Thereafter, some preliminary work in evaluating new spatial reuse options in 802.11ax will be described, along with performance evaluation conducted via the open source network simulator ns-3 (www.nsnam.org). Finally, ongoing work on additional features (OFDMA and Uplink Orthogonal Multiple Access) will be mentioned along with new research needed to enable spectrum sharing with 5G NR Unlicensed (5G NR-U) in the 6 GHz band.

Smart environments, such as smart cities and smart homes, are Cyber-Physical-Systems (CPSs) which are becoming an increasing part of our everyday lives. Several applications in these systems, such as energy management through home appliance identification or activity recognition, adopt Machine Learning (ML) as a practical tool for extracting useful knowledge from raw data. These applications are usually characterized by a sequential stream of data, unlike the classical ML scenario in which the entire data is available during training. For such applications, Stream-based Active Learning (SAL) has been designed as a type of supervised ML in which an expert is asked to label the most informative instances as they arrive. Previous SAL techniques assume that the expert is always available and always labels the data correctly. However, in several applications, such as those mentioned above, the SAL activity interweaves with the everyday life of regular residents, who are often not experts, and may also not always be willing to participate in the labeling process. In this paper, we discuss the importance of user-centered ML, and show how taking into account realistic models of user behavior significantly improves the accuracy and reduces the training period of smart environment applications based on SAL. We consider two use cases, namely appliance identification and activity recognition. Results based on real data sets show an improvement in terms of accuracy up to 55.38%.

This paper deals with techniques to reduce the reliance of IP address-lookup schemes in packet-switched routers on Ternary Content Associative Memory (TCAM), as is commonly implemented in current-generation IP routers. The proposed approaches involve cooperation among neighboring routers, where some form of route lookup clue is provided a router to its downstream router. This is used to reduce the address lookup time, which is particularly important is routers handling links of 100-400 Gbps. Four schemes are compared, of which three are proposed in this paper and are called Lookup at Egress (LE), Unique Memory Index (UM) and flow index (FI) schemes. They respectively provide egress port, global and local memory indices as lookup clues to their next-hop routers. A P4 switch-based implementation has been used to establish the feasibility of the proposed schemes. Further, a detailed performance study shows that up to 80% of the TCAM lookups can be eliminated by these schemes.