Vision transformer (ViTs) models have shown higher accuracy, robustness and large volume data processing ability, creating new baselines and references for perception tasks. However, these advantages require large memory and high-performance processors and computing units, which makes model adaptability and deployment challenging within resource-constrained environments such as memory-restricted and battery-powered edge devices. This paper addresses the model deployment challenges by proposing a model approximation approach VI-ViT, for edge deployment using variational inference with mixed precision for processing multi-modalities, such as point clouds and images. Our experimental evaluation on the nuScenes and Waymo datasets show up to 37% and 31% reduction in model parameters and Flops while maintaining a mean average precision of 70.5 compared to 74.8 of the baseline model. This work presents a practical deployment approach for approximating and optimizing Vision Transformers for edge AI applications by balancing model metrics such as parameters, flops, latency, energy consumption, and accuracy, which can easily be adapted to other transformer models and datasets.

Driving assist applications and connected autonomous vehicle systems are supported using AI models and algorithms, which process and analyze heavy data volumes. High-performance computing units and large memory systems support these models, algorithms, and applications, which results in additional onboard energy consumption. The current trend is also towards full electrification of vehicles and increasing connectivity in the vehicular ecosystem to support collaborative and distributed applications using vehicle-edge-cloud computing. However, with the increased focus on model performance and improving the accuracy of these models and applications, the issue of high-performance computing requirements and resulting energy consumption are overlooked. The problem becomes more challenging and complex for resource-constrained edge devices, which are battery-dependent and have limited memory and computing power. This paper proposes components for an adaptive framework to reduce energy consumption by balancing model accuracy. The contributions include proposing and integrating model partition mechanisms, adaptive deployment across edge devices and approximation strategies for the models. By integrating these components, this framework supports energy-aware development across various platforms. The approach offers a sustainable method for computing and communication-oriented applications within the vehicular ecosystem.

Vulnerable road users (VRUs), including pedestrians, cyclists, and motorcyclists, account for approximately 50% of road traffic fatalities globally, as per the World Health Organization. In these scenarios, the accuracy and fairness of perception applications used in autonomous driving become critical to reduce such risks. For machine learning models, performing object classification and detection tasks, the focus has been on improving accuracy and enhancing model performance metrics; however, issues such as biases inherited in models, statistical imbalances and disparities within the datasets are often overlooked. Our research addresses these issues by exploring class imbalances among vulnerable road users by focusing on class distribution analysis, evaluating model performance, and bias impact assessment. Using popular CNN models and Vision Transformers (ViTs) with the nuScenes dataset, our performance evaluation shows detection disparities for underrepresented classes. Compared to related work, we focus on metric-specific and cost-sensitive learning for model optimization and bias mitigation, which includes data augmentation and resampling. Using the proposed mitigation approaches, we see improvement in IoU(%) and NDS(%) metrics from 71.3 to 75.6 and 80.6 to 83.7 for the CNN model. Similarly, for ViT, we observe improvement in IoU and NDS metrics from 74.9 to 79.2 and 83.8 to 87.1. This research contributes to developing reliable models while addressing inclusiveness for minority classes in datasets. Code can be accessed at: BiasDet.

Deploying scalable Vision Transformer applications on mobile and edge devices is constrained by limited memory and computational resources. Existing model development and deployment strategies include distributed computing and inference methods such as federated learning, split computing, collaborative inference and edge-cloud offloading mechanisms. While these strategies have deployment advantages, they fail to optimize memory usage and processing efficiency, resulting in increased energy consumption. This paper optimizes energy consumption by introducing adaptive model partitioning mechanisms and dynamic scaling methods for ViTs such as EfficientViT and TinyViT, adjusting model complexity based on the available computational resources and operating conditions. We implement energy-efficient strategies that minimize inter-layer communication for distributed machine learning across edge devices, thereby reducing energy consumption from data flow and computation. Our evaluations on a series of benchmark models show improvements, including up to a 32.6% reduction in latency and 16.6% energy savings, while maintaining mean average precision sacrifices within 2.5 to 4.5% of baseline models. These results show that our proposal is a practical approach for improving edge AI sustainability and efficiency.

Edge artificial intelligence (AI) is an innovative computing paradigm that aims to shift the training and inference of machine learning models to the edge of the network. This paradigm offers the opportunity to significantly impact our everyday lives with new services such as autonomous driving and ubiquitous personalized health care. Nevertheless, bringing intelligence to the edge involves several major challenges, which include the need to constrain model architecture designs, the secure distribution and execution of the trained models, and the substantial network load required to distribute the models and data collected for training. In this article, we highlight key aspects in the development of edge AI in the past and connect them to current challenges. This article aims to identify research opportunities for edge AI, relevant to bring together the research in the fields of artificial intelligence and edge computing.

HTTP/3 (H3) has experienced significant growth and extensive adoption in various scenarios, especially in Content Delivery Networks (CDNs). Over the past few years, there have been numerous insightful studies on its deployment in industrial CDNs. However, these studies often separately analyze H3 and CDN, overlooking their synergistic integration. In this work, we explore the applicability of H3 in CDN from a holistic perspective. We analyze 325 websites hosted by seven CDN providers and identify three key characteristics where CDN align perfectly with H3's strengths. Firstly, CDN resources dominate the composition of webpages, where enabling H3 can amplify H3's benefits in connection acceleration. Secondly, CDN providers also exhibit a dominant characteristic, with the majority of CDN resources hosted by a few large providers. This phenomenon makes different webpages share the same provider. When browsing consecutively, H3 helps to skip the connection phase by resuming the connections to the same CDN provider across pages. Thirdly, H3 mitigates the congestion problem on webpages serving multiple CDN resources. This work provides a deeper insight into the applicability of H3 in large-scale distributed systems like CDNs, holding promise for informing the development and optimization of industrial H3.

We demonstrate SPATIAL, a proof-of-concept system that augments modern applications with capabilities to analyze trustworthy properties of AI models. The practical analysis of trustworthy properties is key to guaranteeing the safety of users and overall society when interacting with AI -driven applications. SPATIAL implements AI dashboards to introduce human-in-the-loop capabilities for the construction of AI models. SPATIAL allows different stakeholders to obtain quantifiable insights that characterize the decision making process of AI. This information can then be used by the stakeholders to comprehend possible issues that influence the performance of AI models, such that the issues can be resolved by human operators. Through rigorous benchmarks and experiments in a real-world industrial application, we demonstrate that SPATIAL can easily augment modern applications with metrics to gauge and monitor trustworthiness. However, this, in turn, increases the complexity of developing and maintaining the systems implementing AI. Our work paves the way towards augmenting modern applications with trustworthy AI mechanisms and human oversight approaches.

Digitalization and datafication of financial systems result in more efficiency, but might also result in the exclusions of certain groups. Governments are looking for ways to increase inclusions and leave no one behind. For this, they must govern an organizational ecosystem of public and private parties. We derive value-based requirements through a systematic research methodology and iteratively refine design principles for achieving inclusivity goals. This refinement process is enriched by interviews with field experts, leading to the formulation of key Design principles: the essential role of inclusive metrics, leveraging alternative data sources, ensuring transparency in loan processes and the ability for decision contestation, providing tailored credit solutions, and maintaining long-term system sustainability. The government's role is to ensure a level playing field where all parties have equal access to the data. Following the principles ensures that exclusion and discrimination become visible and can be avoided. This study underscores the necessity for system-level transformations, inclusion-by-design, and advocacy for a new system design complemented by regulatory updates, new data integration, inclusive AI, and organizational collaborative shifts. These principles can also be used in different data-driven governance situations.

Our democratic systems have been challenged by the proliferation of artificial intelligence (AI) and its pervasive usage in our society. For instance, by analyzing individuals’ social media data, AI algorithms may develop detailed user profiles that capture individuals’ specific interests and susceptibilities. These profiles are leveraged to derive personalized propaganda, with the aim of influencing individuals toward specific political opinions. To address this challenge, the value of privacy can serve as a bridge, as having a sense of privacy can create space for people to reflect on their own political stance prior to making critical decisions, such as voting for an election. In this paper, we explore a novel approach by harnessing the potential of AI to enhance the privacy of social-media data. By leveraging adversarial machine learning, i.e., “AI versus AI,” we aim to fool AI-generated user profiles to help users hold a stake in resisting political profiling and preserve the deliberative nature of their political choices. More specifically, our approach probes the conceptual possibility of infusing people’s social media data with minor alterations that can disturb user profiling, thereby reducing the efficacy of the personalized influences generated by political actors. Our study delineates the boundary of ethical and practical implications associated with this ‘AI versus AI’ approach, highlighting the factors for the AI and ethics community to consider in facilitating deliberative decision-making toward democratic elections.

The temporal patterns of code submissions, denoted as work rhythms, provide valuable insight into the work habits and productivity in software development. In this paper, we investigate the work rhythms in software development and their effects on technical performance by analyzing the profiles of developers and projects from 110 international organizations and their commit activities on GitHub. Using clustering, we identify four work rhythms among individual developers and three work rhythms among software projects. Strong correlations are found between work rhythms and work regions, seniority, and collaboration roles. We then define practical measures for technical performance and examine the effects of different work rhythms on them. Our findings suggest that moderate overtime is related to good technical performance, whereas fixed office hours are associated with receiving less attention. Furthermore, we survey 92 developers to understand their experience with working overtime and the reasons behind it. The survey reveals that developers often work longer than required. A positive attitude towards extended working hours is associated with situations that require addressing unexpected issues or when clear incentives are provided. In addition to the insights from our quantitative and qualitative studies, this work sheds light on tangible measures for both software companies and individual developers to improve the recruitment process, project planning, and productivity assessment.

An up-to-date and comprehensive guide to mobile edge computing and communications Mobile Edge Computing and Communications offers a practical guide to mobile edge computing and communications (MEC). With contributions from noted experts on the topic, the book covers the design, deployment, and operational aspects of this rapidly growing domain. The text provides the information needed to understand the mainstream system architectures and integration methods that have been proposed in MEC. In addition, the book clearly illustrates critical lifecycle functions and stages of MEC, and shows how to deploy MEC in 5G and beyond mobile networks. Comprehensive in scope, the book contains discussions on the challenges and opportunities of mobile edge computing and communications concepts combined with the most relevant emerging applications and services. The authors provide insights for all relative stakeholders of mobile networks such as mobile network operators. This important book: • Provides a comprehensive walkthrough of mobile edge computing and communications; • Includes detailed analysis of current edge applications and technology foundation; • Presents information on driving forces and future directions of MEC; • Provides an authentic source of information from industry experts to drive the future of computing. Written for mobile network operators, ICT service developers, academic researchers, undergraduate and graduate students, Mobile Edge Computing and Communications offers a guide to the current and future of MEC that will enable a completely new paradigm for future computing and communications.

Recent advancements in hardware and software systems have been driven by the deployment of emerging smart health and mobility applications. These developments have modernized the traditional approaches by replacing conventional computing systems with cyber–physical and intelligent systems combining the Internet of Things (IoT) with Edge Artificial Intelligence. Despite the many advantages and opportunities of these systems within various application domains, the scarcity of energy, extensive computing needs, and limited communication must be considered when orchestrating their deployment. Inducing savings in these directions is central to the Approximate Computing (AxC) paradigm, in which the accuracy of some operations is traded off with energy, latency, and/or communication reductions. Unfortunately, the dynamics of the environments in which AxC-equipped IoT systems operate have been paid little attention. We bridge this gap by surveying adaptive AxC techniques applied to three emerging application domains, namely autonomous driving, smart sensing and wearables, and positioning, paying special attention to hardware acceleration. We discuss the challenges of such applications, how adaptive AxC can aid their deployment, and which savings it can bring based on traits of the data and devices involved. Insights arising thereof may serve as inspiration to researchers, engineers, and students active within the considered domains.

In recent years, there has been a notable increase in the size of commonly used image classification models. This growth has empowered models to recognize thousands of diverse object types. However, their computational demands pose significant challenges, especially when deploying them on resource-constrained edge devices. In many use cases where a model is deployed on an edge device, only a small subset of the classes will ever be observed by a given model instance. Our proposed test-time specialization of dynamic neural networks allows these models to become faster at recognizing the classes that are observed frequently, while maintaining the ability to recognize all other classes, albeit slightly less efficient. We benchmark our approach on a real-world edge device, obtaining significant speedups compared to the baseline model without test-time adaptation.

Model partitioning is a promising solution to reduce the high computation load and transmission of high-volume data. Within the scope of Edge AI, the fundamentals of model partitioning involve splitting the model for local computing at the edge and offloading heavy computation tasks to the cloud or server. This approach benefits scenarios with limited computing and battery capacity with low latency requirements, such as connected autonomous vehicles. However, while model partitioning offers advantages in reducing the onboard computation, memory requirements and inference time, it also introduces challenges such as increased energy consumption for partitioned computations and overhead for transferring partitioned data/model. In this work, we explore hybrid model partitioning to optimize computational and communication energy consumption. Our results provide an initial analysis of the tradeoff between energy and accuracy, focusing on the energy-aware model partitioning for future Edge AI applications.

Organizations are increasingly reliant on third-party software products to expedite their own development cycles, often incorporating numerous components into their end systems, resulting in a lack of transparency in software dependencies. Malicious actors exploit this, leading to Software Supply Chain (SSC) attacks with substantial economic and security damages. To mitigate this threat, the Software Bill of Materials (SBOM) concept was introduced. It details software components and their supply chain relationships, thus enhancing SSC transparency. Unfortunately, SBOM adoption still remains limited. While previous studies identified some reasons behind this, they overlooked the perspectives of different business stakeholder groups involved in SBOM's lifecycle.

In this work, we address this gap by studying business stakeholder groups directly involved in SBOM production and consumption. The main goal of this work is to identify which groups can drive or inhibit SBOM adoption and the rationale behind this behavior. By conducting interviews with the group representatives, we identified stakeholder-specific risks, benefits, concerns and incentives regarding SBOM adoption. Our analysis suggests that SBOM adoption potential is higher among System Integrators and Software Vendors. At the same time, B2B customers and Individual Developers have the least motivation, inhibiting the process of SBOM adoption. Given that these are the main SBOM consuming and supplying stakeholders correspondingly, we conclude that the overall adoption potential of this technology is currently limited and requires considerable external impulse.

Despite its enormous economical and societal impact, lack of human-perceived control and safety is re-defining the design and development of emerging AI-based technologies. New regulatory requirements mandate increased human control and oversight of AI, transforming the development practices and responsibilities of individuals interacting with AI. In this paper, we present the SPATIAL architecture, a system that augments modern applications with capabilities to gauge and monitor trustworthy properties of AI inference capabilities. To design SPATIAL, we first explore the evolution of modern system architectures and how AI components and pipelines are integrated. With this information, we then develop a proof-of- concept architecture that analyzes AI models in a human-in-the- loop manner. SPATIAL provides an AI dashboard for allowing individuals interacting with applications to obtain quantifiable insights about the AI decision process. This information is then used by human operators to comprehend possible issues that influence the performance of AI models and adjust or counter them. Through rigorous benchmarks and experiments in real- world industrial applications, we demonstrate that SPATIAL can easily augment modern applications with metrics to gauge and monitor trustworthiness, however, this in turn increases the complexity of developing and maintaining systems implementing AI. Our work highlights lessons learned and experiences from augmenting modern applications with mechanisms that support regulatory compliance of AI. In addition, we also present a road map of on-going challenges that require attention to achieve robust trustworthy analysis of AI and greater engagement of human oversight.

Outstanding users (OUs) denote the influential, 'core' or 'bridge' users in online social networks. How to accurately detect and rank them is an important problem for third-party online service providers and researchers. Conventional efforts, ranging from early graph-based algorithms to recent machine learning-based approaches, typically rely on an entire social network's information. However, for privacy-conscious users or newly-registered users, such information is not easily accessible. To address this issue, we present DeepPick, a novel framework that considers both the generalization and specialization in the detection task of OUs. For generalization, we introduce deep neural networks to capture dynamic features of the users. For specialization, we leverage the traditional descriptive features to make use of public information about users. Extensive experiments based on real-world datasets demonstrate that our approach achieves a high efficacy of detection performance against the state-of-the-art.

Billions of distributed, heterogeneous, and resource constrained IoT devices deploy on-device machine learning (ML) for private, fast, and offline inference on personal data. On-device ML is highly context dependent and sensitive to user, usage, hardware, and environment attributes. This sensitivity and the propensity toward bias in ML makes it important to study bias in on-device settings. Our study is one of the first investigations of bias in this emerging domain and lays important foundations for building fairer on-device ML. We apply a software engineering lens, investigating the propagation of bias through design choices in on-device ML workflows. We first identify reliability bias as a source of unfairness and propose a measure to quantify it. We then conduct empirical experiments for a keyword spotting task to show how complex and interacting technical design choices amplify and propagate reliability bias. Our results validate that design choices made during model training, like the sample rate and input feature type, and choices made to optimize models, like light-weight architectures, the pruning learning rate, and pruning sparsity, can result in disparate predictive performance across male and female groups. Based on our findings, we suggest low effort strategies for engineers to mitigate bias in on-device ML.

The smishing-based malware Flubot was taken down in mid-2022, yet there is little understanding of how it directly impacted smartphone users. We engage with customers of a partner Internet Service Provider (ISP), who have suffered a Flubot infection on their smartphones. We surveyed 87 ISP customers who had been notified of a Flubot infection, in the months around and preceding the take-down of Flubot. We found that slightly over half of respondents were unaware of the malware infection before being notified, though many others had suspicions. We also observe that just over half of respondents experienced non-technical harms from the malware, with many experiencing harms before notification and several experiencing unwanted or aggressive activity from users of other infected devices. Many respondents reported not having removed the malware, while some discarded the infected device or stopped using online services in their efforts to be more secure afterwards. We offer recommendations, including that clearer guidance be sought to help users identify a malware infection (and not a focus only on prevention), and support provided for recovery from personal harms caused by mobile malware, as the impacts are not only technical.