Feasibility of Longest Prefix Matching using Learned Index Structures

This paper revisits longest prefix matching in IP packet forwarding because an emerging data structure, learned index, is recently presented. A learned index uses machine learning to associate key-value pairs in a key-value store. The fundamental idea to apply a learned index to an FIB is to simplify the complex longest prefix matching operation to a nearest address search operation. The size of the proposed FIB is less than half of an existing trie-based FIB while it achieves the computation speed nearly equal to the trie-based FIB. Moreover, the computation speed of the proposal is independent of the length of IP prefixes, unlike trie-based FIBs.

What's my App?

With the spread of broadband Internet, Real-Time Communication (RTC) platforms have become increasingly popular and have transformed the way people communicate. Thus, it is fundamental that the network adopts traffic management policies that ensure appropriate Quality of Experience to users of RTC applications. A key step for this is the identification of the applications behind RTC traffic, which in turn allows to allocate adequate resources and make decisions based on the specific application's requirements. In this paper, we introduce a machine learning-based system for identifying the traffic of RTC applications. It builds on the domains contacted before starting a call and leverages techniques from Natural Language Processing (NLP) to build meaningful features. Our system works in real-time and is robust to the peculiarities of the RTP implementations of different applications, since it uses only control traffic. Experimental results show that our approach classifies 5 well-known meeting applications with an F1 score of 0.89.

Improving Web QoE Monitoring for Encrypted Network Traffic through Time Series Modeling

This paper addresses the problem of Quality of Experience (QoE) monitoring for web browsing. In particular, the inference of common Web QoE metrics such as Speed Index (SI) is investigated. Based on a large dataset collected with open web-measurement platforms on different device-types, a unique feature set is designed and used to estimate the RUMSI - an efficient approximation to SI, with machinelearning based regression and classification approaches. Results indicate that it is possible to estimate the RUMSI accurately, and that in particular, recurrent neural networks are highly suitable for the task, as they capture the network dynamics more precisely.

Cheating at Craps

Craps is a simple dice game that is popular in casinos around the world. While the rules for Craps, and its mathematical analysis, are reasonably straightforward, this paper instead focuses on the best ways to cheat at Craps, by using loaded (biased) dice. We use both analytical modeling and simulation modeling to study this intriguing dice game. Our modeling results show that biasing a die away from the value 1 or towards the value 5 lead to the best (and least detectable) cheating strategies, and that modest bias on two loaded dice can increase the winning probability above 50%. Our Monte Carlo simulation results provide validation for our analytical model, and also facilitate the quantitative evaluation of other scenarios, such as heterogeneous or correlated dice.

Keynote

Blockchain is a novel distributed ledger technology. Through its features and smart contract capabilities, a wide range of application areas opened up for blockchain-based innovation [5]. In order to analyse how concrete blockchain systems as well as blockchain applications are used, data must be extracted from these systems. Due to various complexities inherent in blockchain, the question how to interpret such data is non-trivial. Such interpretation should often be shared among parties, e.g., if they collaborate via a blockchain. To this end, we devised an approach codify the interpretation of blockchain data, to extract data from blockchains accordingly, and to output it in suitable formats [1, 2]. This work will be the main topic of the keynote. In addition, application developers and users of blockchain applications may want to estimate the cost of using or operating a blockchain application. In the keynote, I will also discuss our cost estimation method [3, 4]. This method was designed for the Ethereum blockchain platform, where cost also relates to transaction complexity, and therefore also to system throughput.

Blockchain Privacy Through Merge Avoidance and Mixing Services

Cryptocurrencies typically aim at preserving the privacy of their users. Different cryptocurrencies preserve privacy at various levels, some of them requiring users to rely on strategies to raise the privacy level to their needs. Among those strategies, we focus on two of them: merge avoidance and mixing services. Such strategies may be adopted on top of virtually any blockchain-based cryptocurrency. In this paper, we show that whereas optimal merge avoidance leads to an NP-hard optimization problem, incentive-compatible mixing services are subject to a certain class of impossibility results. Together, our results contribute to the body of work on fundamental limits of privacy mechanisms in blockchainbased cryptocurrencies.

Machine Learning based KPI Monitoring of Video Streaming Traffic for QoE Estimation

Quality of Experience (QoE) monitoring of video streaming traffic is crucial task for service providers. Nowadays it is challenging due to the increased usage of end-to-end encryption. In order to overcome this issue, machine learning (ML) approaches for QoE monitoring have gained popularity in the recent years. This work proposes a framework which includes a machine learning pipeline that can be used for detecting key QoE related events such as buffering events and video resolution changes for ongoing YouTube video streaming sessions in real-time. For this purpose, a ML model has been trained using YouTube streaming traffic collected from Android devices. Later on, the trained ML model is deployed in the framework's pipeline to make online predictions. The ML model uses statistical traffic information observed from the network-layer for learning and predicting the video QoE related events. It reaches 88% overall testing accuracy for predicting the video events. Although our work is yet at an early stage, the application of the ML model for online detection and prediction of video events yields quite promising results.

Crypto-Hotwire

Blockchains and cryptocurrencies disrupted the conversion of energy into a medium of exchange. Numerous applications for blockchains and cryptocurrencies are now envisioned for purposes ranging from inventory control to banking applications. Naturally, in order to mine in an economically viable way, regions where energy is plentiful and cheap, e.g., close to hydroelectric plants, are sought. The possibility of converting energy into cash, however, also opens up opportunities for a new kind of cyber attack aimed at illegally mining cryptocurrencies by stealing energy. In this work, we indicate, using data from January and February of 2018 from our university, that such a threat is real, and present a projection of the gains derived from these attacks.

Robustness analysis of Bitcoin confirmation times

Bitcoin payments require a random amount of time to get confirmed (i.e. to be grouped by the miners into a block and to be added to the Bitcoin blockchain). In [8, 11], the authors propose the modelling of the Bitcoin confirmation time by the so-called time to ruin of the Cramer-Lundberg (CL) model. This provides off-the-shelf results directly aimed at predicting the confirmation time. However, analyses suggest that the data may not fully conform with the CL model assumptions. In this manuscript, we show by means of a robustness analysis that the time to ruin of a CL model is near insensitive to small changes in the model assumptions and illustrate that the proposed heuristic model can be used to accurately predict the confirmation times even when the data deviate (to a small degree) from the model assumptions.

Mobile Task Offloading Under Unreliable Edge Performance

Offloading resource-hungry tasks from mobile devices to an edge server has been explored recently to improve task com- pletion time as well as save battery energy. The low la- tency computing resource from edge servers are a perfect companion to realize such task offloading. However, edge servers may su er from unreliable performance due to its rapid workload variation and reliance on intermittent re- newable energy. Further, batteries in mobile devices make online optimum offloading decisions challenging since it in- tertwines offloading decisions across di erent tasks. In this paper, we propose a deep Q-learning based task offloading solution, DeepTO, for online task offloading. DeepTO learns edge server performance in a model-free manner and takes future battery needs of the mobile device into account. Us- ing a simulation-based evaluation, we show that DeepTO can perform close to the optimum solution that has com- plete future knowledge.