The widespread deployment of machine learning systems in critical real-world decision-making applications has highlighted the urgent need for counterfactual explainability methods that operate effectively. Global counterfactual explanations, expressed as actions to offer recourse, aim to provide succinct explanations and insights applicable to large population subgroups. High effectiveness, measured by the fraction of the population that is provided recourse, ensures that the actions benefit as many individuals as possible. Keeping the cost of actions low ensures the proposed recourse actions remain practical and actionable. Limiting the number of actions that provide global counterfactuals is essential to maximize interpretability. The primary challenge, therefore, is to balance these trade-offs—maximizing effectiveness, minimizing cost, while maintaining a small number of actions. We introduce GLANCE, a versatile and adaptive algorithm that employs a novel agglomerative approach, jointly considering both the feature space and the space of counterfactual actions, thereby accounting for the distribution of points in a way that aligns with the model's structure. This design enables the careful balancing of the trade-offs among the three key objectives, with the size objective functioning as a tunable parameter to keep the actions few and easy to interpret. Our extensive experimental evaluation demonstrates that GLANCE consistently shows greater robustness and performance compared to existing methods across various datasets and models.

Serverless computing has gained increasing interest in recent years for enabling large-scale machine learning tasks. However, training a machine learning model in a serverless setting is a complex task and several challenges need to be addressed particularly in data distribution, result aggregation, resource heterogeneity, failures, container ephemerality, network and execution cost. These difficulties stem from the inherent complexity of distributed computation and the coordination demands of the machine learning algorithms. We propose STRATA2.0, a serverless middleware for Machine Learning training in serverless environments. STRATA2.0 provides a comprehensive suite of mechanisms designed to support efficient training of machine learning models on serverless infrastructures and address key challenges related to efficient data communication, coordination and synchronization, scalable and time efficient training of ML models using heterogeneous containers. Our extensive experimental results demonstrate that STRATA2.0 achieves the same level of accuracy with fewer data points, is on average three times faster in training time compared to centralized approaches, reduces energy consumption by up to 50%, and remains resilient when up to 60% of training instances fail.

The proliferation of smart technologies has provided significant advances in the way people receive information and interact in smart cities. For instance, smart traffic monitoring systems allow citizens and city operators to receive in real-time alerts about traffic conditions. At the same time, alternative means of transport, such as bike sharing systems, have enjoyed tremendous success in many major cities around the world today and can provide real-time information regarding the mobility of the users. It is quite clear that smart cities systems generate vast amounts of data during their daily operation. Researchers typically focus on analyzing such urban data in order to identify mobility patterns or infer user preferences to provide services to the crowd, such as traffic warning applications. In this thesis, we aimed at analyzing the vast amounts of urban data generated in these systems, that are large in volume, and typically heterogeneous, to identify key insights in their operation and develop a suite of crowdsourcing techniques to improve the working of smart city applications and ultimately enhance the sustainability of smart cities.

The radical advances in mobile computing, the IoT technological evolution along with cyberphysical components (e.g., sensors, actuators, control centers) have led to the development of smart city applications that generate raw or preprocessed data, enabling workflows involving the city to better sense the urban environment and support citizens’ everyday lives. Recently, a new era of Mobile Edge Cloud (MEC) infrastructures has emerged to support smart city applications that aim to address the challenges raised due to the spatio-temporal dynamics of the urban crowd as well as bring scalability and on-demand computing capacity to urban system applications for timely response. In these, resource capabilities are distributed at the edge of the network and in close proximity to end-users, making it possible to perform computation and data processing at the network edge. However, there are important challenges related to real-time execution, not only due to the highly dynamic and transient crowd, the bursty and highly unpredictable amount of requests but also due to the resource constraints imposed by the Mobile Edge Cloud environment. In this paper, we present TIMBER, our framework for efficiently supporting mobile daTa processing pIpelines in MoBile cloud EnviRonments that effectively addresses the aforementioned challenges. Our detailed experimental results illustrate that our approach can reduce the operating costs by 66.245% on average and achieve up to 96.4% similar throughput performance for agnostic workloads.

Serverless computing has received growing interest in recent years for supporting large-scale machine learning tasks. However, training a machine learning model in a serverless environment is a nontrivial procedure and several challenges still need to be addressed in the data distribution and result aggregation steps as well as the cost of execution due to the inherent complexity of the distributed computation and the coordination required in the learning algorithm. In this work, we focus on Random Forests, a state-of-the-art technique in many Machine Learning applications. We propose STRATA, a cost-effective middleware to train Random Forests atop a serverless environment that successfully addresses these training challenges. As we show in our extensive experimental evaluation STRATA achieves 3X better training times on average compared to a centralized approach and can withstand up to 70% of failures during training.

The proliferation of smartphone devices has led to the emergence of powerful user services from enabling interactions with friends and business associates to mapping, finding nearby businesses and alerting users in real-time. Moreover, users do not realize that continuously sharing their trajectory data with online systems may end up revealing a great amount of information in terms of their behavior, mobility patterns and social relationships. Thus, addressing these privacy risks is a fundamental challenge. In this work, we present$TP^{3}$, a Privacy Protection system for Trajectory analytics. Our contributions are the following: (1) we model a new type of attack, namely “social link exploitation attack”, (2) we utilize the coresets theory, a fast and accurate technique which approximates well the original data using a small data set, and running queries on the coreset produces similar results to the original data, and (3) we employ the Serverless computing paradigm to accommodate a set of privacy operations for achieving high system performance with minimized provisioning costs, while preserving the users' privacy. We have developed these techniques in our$TP^{3}$system that works with state-of-the-art trajectory analytics apps and applies different types of privacy operations. Our detailed experimental evaluation illustrates that our approach is both efficient and practical.

The problem of privacy protection of trajectory data has received increasing attention in recent years with the significant grow in the volume of users that contribute trajectory data with rich user information. This creates serious privacy concerns as exposing an individual's privacy information may result in attacks threatening the user's safety. In this demonstration we present$TP^{3}$a novel practical framework for supporting trajectory privacy preservation in Mobile Cloud Environments (MCEs). In$TP^{3}$, non-expert users submit their trajectories and the system is responsible to determine their privacy exposure before sharing them to data analysts in return for various benefits, e.g. better recommendations.$TP^{3}$makes a number of contributions: (a) It evaluates the privacy exposure of the users utilizing various privacy operations, (b) it is latency-efficient as it implements the privacy operations as serverless functions which can scale automatically to serve an increasing number of users with low latency, and (c) it is practical and cost-efficient as it exploits the serverless model to adapt to the demands of the users with low operational costs for the service provider. Finally,$TP^{3}$'s Web-UI provides insights to the service provider regarding the performance and the respective revenue from the service usage, while enabling the user to submit the trajectories with recommended preferences of privacy.

In recent years we have witnessed a growing interest in trip planning systems aiming at organizing daily travel schedules in smart cities. Such systems use specialized engines to find optimal means of transport between two geospatial endpoints to provide recommendations to citizens for short routes across the city. At the same time, alternative means of transportation, such as bike sharing systems, have enjoyed tremendous success since they offer a green and facile solution for daily commuters and tourists. However, one major challenge of the bike sharing systems is that the distribution of bikes among the stations can be quite uneven during rush hours or due to topography. This often results in shortage of bikes and increasing numbers of disappointed users. Existing works in the literature are limited since they only focus on predicting the demand or apply a-posteriori methods for balancing the load of stations. Furthermore, none of these works consider the benefit of these systems in concert. In this work, we present "MOToR" (MultimOdal Trip Rebalancing), a system that builds upon the OpenTripPlanner framework to incorporate dynamic transit schedule data while balancing the availability of bikes among the bike stations. Our experimental evaluation shows that our approach is practical, efficient and outperforms state-of-the-art methods for route planning.

Bike-sharing systems have enjoyed tremendous success in many major cities around the world today as a new means of urban public transportation offering a green and facile solution for daily commuters and tourists. One common problem featured in these systems is that the distribution of bikes among stations can be quite uneven, due to topography, rush hours or during the occurrence of major events around the city. This often results in shortage of bikes or bike parking racks. An unbalanced bike system means an unreliable form of transportation and disappointed users. Existing works in the literature are limited as they are not designed to handle fluctuating, high or unpredictable demand during large city events that typically affect multiple stations and require rebalancing in real-time, during the event, to ensure seamless operation. In this work, we present “SmartBIKER”, a cost-effective framework for bike sharing systems focusing on major city events. SmartBIKER models bike demand trends during major events, identifies bike stations with low or high demand using a trend forecasting model and determines a relocation strategy that minimizes the relocation cost while maximizing the utility of the stations. Our experimental evaluation shows that our approach is practical, efficient and outperforms state-of-the-art relocation and prediction schemes.

In recent years we are witnessing a growing interest in identifying various aspects affecting the quality of life in smart cities, such as traffic congestion and pollution levels, in order to provide services that enhance the public welfare. In smart cities, sensor infrastructures are deployed around the city combined with data analytics, to monitor and detect in real-time possible anomalies or events of interest. One major challenge that arise in smart-cities is to evaluate the health state of an urban city using heterogeneous multi-source urban data, i.e., pollution and traffic data. Existing works in the literature are limited since they analyze a single source of data, either inferring the air quality or estimating traffic congestion. However, none of these works considers both data sources in concert for estimating the city's health state. In this work, we present “HELIoS” (HEalthy LIving Smart), a framework that combines multiple heterogeneous sources of data, i.e., urban traffic and pollution data, to diagnose the health state of urban areas in a smart city. Our experimental evaluation provides valuable insights into identifying the health state of an urban area, and shows that our approach is both practical and efficient.

Recommending nearby Points of Interest (POI) has received growing interest in mobile location-based networks today, where users share content embedded with location information. In this work, we propose a novel caching framework to support personalised proactive caching for mobile location-based social networks. We propose "LOCAI", which uses a probabilistic approach in order to predict the POIs that users will access and retrieve the appropriate data objects that will fulfill user preferences. Our detailed experimental evaluation, using data from the Foursquare location-based social network, illustrates that LOCAI minimizes the user latency to retrieve the data objects they are interested in, is efficient and practical.