Finding a maximum-weight matching is a classical and well-studied problem in computer science, solvable in cubic time in general graphs. We consider the specialization called assignment problem where the input is a bipartite graph, and introduce in this work the “discovery” variant considering edge weights that are not provided as input but must be queried , requiring additional and costly computations. We develop discovery algorithms here to minimize the number of queried weights while providing guarantees on the computed solution. In this work, we first show the inherent challenges of designing discovery algorithms for general assignment problems. We then provide and analyze several efficient greedy algorithms that can make use of natural assumptions about the order in which the nodes are processed by the algorithms. Our motivations for exploring this problem stem from finding practical solutions to a variation of maximum weight matching in bipartite hypergraphs, a problem recently emerging in the formation of peer-to-peer energy-sharing communities.

Efficient monitoring of large distributed systems is critical for applications such as data center load balancing, fleet management, and smart grid energy optimization. This paper addresses the continuous distributed monitoring problem, where a central coordinator tracks statistics from numerous distributed nodes in real time. We present FEDAMON, a novel F orecast-based, E rror-bounded, and D ata- A ware approach to continuous distributed Mon itoring that significantly reduces communication costs while maintaining accuracy. Instead of transmitting all values to the coordinator, our event-based monitoring leverages lightweight forecasting models at the coordinator and distributed nodes to predict the evolution of observations, communicating when deviations exceed an error threshold. To adapt to dynamically changing data streams, we introduce a data-aware model selection strategy that optimizes the trade-off between communication and accuracy. Our solution is communication-efficient, fully automated, and equipped with dynamic error control, reducing system parametrization to a single error tolerance of three preset levels. FEDAMON reduces communication to 10% of the baseline with less than 2% error across all streams on diverse datasets on average. Moreover, the standard parameter solution surpasses even the best calibrated single models across all error bounds, achieving up to 33% improvement in communication efficiency with identical error guarantees. Further gains of 25% in accuracy is obtained by tuning the data-aware control factor without extra cost. In addition, our framework generalizes effectively to previously unseen datasets. Finally, our dynamic error control achieves comparable performance to fixed bounds. Results highlight the scalability and robustness of FEDAMON, enabling fully automatic, real-time monitoring with large communication savings and marginal error.

The Transparency and Consent Framework (TCF), developed by the Interactive Advertising Bureau (IAB) Europe, provides a de facto standard for requesting, recording, and managing user consent from European end-users. Its goal is to help organizations comply with the General Data Protection Regulation (GDPR) and the ePrivacy Directive (ePD). This framework has previously been found to infringe European data protection law and has subsequently been regularly updated. Previous research on the TCF focused exclusively on web contexts, with no attention given to its implementation in mobile apps. No work has systematically studied the compliance implications of the TCF on Android apps. To address this gap, we investigate the prevalence of the TCF in popular Android apps from the Play Store, and assess whether these apps respect users’ consent banner choices. The TCF introduced minor changes in its new version (v.2.3) on March 1st, 2026, aimed at reducing ambiguity for vendors; these changes do not impact our results. We scraped and downloaded 4482 of the most popular Google Play Store apps on an emulated Android device. We automatically identified that 576 (12.85%) of the 4482 downloadable apps implemented the TCF, and we detected potential legal violations within this subset. By automatically interacting with consent banners, we observed that in 15 (2.6%) of these apps, users’ choices are stored only when consent is granted. Users who refuse consent are shown the consent banner again each time they launch the app. We analyzed the apps’ traffic in two different stages, passive (post interaction with the banner) and active (during banner interaction and post user choices). Network analysis conducted during the passive stage reveals that 66.2% of the analyzed TCF-based apps share personal data through the Google Advertising ID (AAID) without consent – the lawful basis for such processing. Furthermore, 55.3% of apps analyzed during the active stage share AAID before users interact with the apps’ consent banners, violating the prior consent requirement. We further expose concerns regarding Google as the dominant Consent Management Provider (CMP) in our dataset (89.76%), structurally accommodating potential legal violations. Our results suggest that mobile implementations of the TCF are prone to significant non-compliance practices, raising concerns about its effectiveness in helping organizations adhere to legal requirements.

Active learning (AL) is a well-established pedagogical technique that promotes active student participation in traditionally passive settings, such as lectures. Two widely used AL tools in higher education are Kahoot, a game-based platform more commonly associated with K-12 education, and Mentimeter, a tool designed to enhance audience engagement but not originally built for AL. Despite their popularity, these tools are often adopted by universities without thorough evaluation. Our study offers a comparative analysis of Kahoot and Mentimeter, based on several years of teaching experience and data from AL quizzes conducted in a large introductory computer networking course. We examine key features relevant to AL integration, such as tool functionalities, question transitions, presentation formats, and gamification elements. Empirical data from five years and eight course instances includes statistics such as student participation and performance, as well as student feedback. Our findings reveal important differences in tool design and student engagement, highlighting key areas for improvement in both tools to match the needs of our computer and engineering education. We further provide essential recommendations for better use of those AL tools within our computer courses.

Consensus, abstracting myriad problems in which processes must agree on a single value, is one of the most celebrated problems of fault-tolerant distributed computing. Consensus applications include fundamental services for the Cloud and Blockchain environments, and in such challenging environments, malicious behaviors are often modeled as adversarial Byzantine faults. At OPODIS 2010, Mostéfaoui and Raynal (in short, MR) presented a Byzantine-tolerant solution to consensus in which the decided value cannot be proposed only by Byzantine processes. MR has optimal resilience coping with up to t < n / 3 Byzantine nodes over n processes. MR provides this multivalued consensus object (which accepts proposals taken from a finite set of values), assuming the availability of a single binary consensus object (which accepts proposals taken from the set { 0 , 1 } ). This work, which focuses on multivalued consensus, aims to design an even more robust solution than MR. Our proposal expands MR's fault-model with self-stabilization, a vigorous notion of fault-tolerance. In addition to tolerating Byzantine, self-stabilizing systems can automatically recover after arbitrary transient-faults occur. These faults represent any violation of the assumptions according to which the system was designed to operate (provided that the algorithm code remains intact). To the best of our knowledge, we propose the first self-stabilizing solution for multivalued consensus for asynchronous message-passing systems prone to Byzantine failures. Our solution has an O ( t ) stabilization time from arbitrary transient faults.

AI-powered text generation tools have brought about a profound shift in students' approaches to writing assignments, prompting the need for writing supervisors to gain a deeper understanding of how these tools can effectively tackle conventional university assignments. This ongoing work presents a documented and student-centered experiment where such tools are used for crafting the final assignment in a higher education course. Through the lens of a reflective essay, the study provides insights into the experiment's nuances and offers practical considerations that can prove invaluable to writing supervisors. Despite being introduced in late 2022, ChatGPT, a prominent AI language model, has quickly become a focal point in higher education research. This work distinguishes itself from existing literature by presenting a practical guide tailored for writing supervisors involved in overseeing diverse writing processes. By documenting a student's firsthand experience using ChatGPT to generate a self-assessment plan within the context of a university writing supervision course, the study not only explores the benefits and challenges of integrating AI tools but also underscores the importance of responsible usage. In particular, this work furnishes valuable insights for writing supervisors navigating the evolving landscape of AI-driven writing tools, offering a nuanced understanding of their practical implications.

The landscape of higher education experienced a significant transformation due to the COVID-19 pandemic, prompting a swift shift from traditional in-class models to online formats. With the return of courses to on-campus delivery, educators now have a distinct opportunity to contemplate teaching methodologies. Our case study delves into the post-pandemic changes within an introductory and sizable networking course, specifically concentrating on optimizing computer network instruction for first-year engineering students. While prior research has explored lessons learned from the pandemic and the increased adoption of digital education, a notable gap exists in the literature concerning labs, particularly the transition from remote to physical content. The primary objective of the study is to guide a more balanced allocation of student learning time in the post-COVID era, aiming to minimize frustration by refining lab activities for equivalent learning outcomes. To enhance student comprehension of networking protocols, we implemented updates, capitalizing on the shift to campus education and employing three pedagogical approaches: Active Learning (AL), Practice Test/Spaced Practice (PT/SP), and Peer Instructions (PI). These updates were driven by three specific objectives: establishing new physical labs including collaboration between groups to foster PI and AL, conducting beneficial in-class exercises to facilitate live PI and enhance a PT/SP approach, and transitioning to in-class quizzes to evaluate AL between online and in-person lectures. Feedback collected through surveys, statistics from the learning management system, and quiz applications indicates that the updates successfully achieved their objectives. The study offers nuanced reflections and valuable insights into enhancing student understanding of networking protocols in the post-COVID era, emphasizing a balanced distribution of learning time and minimizing frustration associated with lab activities.

In recent years, concerns about the privacy of users data have raised as testified by the wide adoption of the HTTPS protocol over its unencrypted predecessor. This work demonstrates, however, that the encryption used in HTTPS does not guarantee that the user's data is hidden when streaming videos using the DASH protocol. We show that the encryption can be bypassed by exploiting recognizable and predictable patterns produced by DASH in side-channels. To demonstrate our attack, we have collected 100k fingerprints from the SVT Play streaming platform, and have shown that encrypted videos can reliably and quickly be identified by capturing streamed HTTPS traffic and comparing it against the fingerprint database. Compared with previous work, our evaluation demonstrates the superior accuracy in our approach as well as its capacity of swiftly identify videos that are playing from an arbitrary timestamp. Our prototype is, to the best of our knowledge, the fastest and most accurate video streaming recognizer to date, only requiring as little as 12 seconds of network traffic to infer a video title with more than 98% accuracy among a catalogue of 20k videos. Our results call for future updates in the DASH protocol designed to circumvent the privacy leak we have shown. An open-source implementation of our prototype is publicly available at https://github.com/embeage/streaming-identification.

Finding a maximum-weight matching is a classical and well-studied problem in computer science, solvable in cubic time in general graphs. We introduce and consider in this work the “discovery” variant of the bipartite matching problem (or assignment problem) where edge weights are not provided as input but must be queried, requiring additional and costly computations. Hence, discovery algorithms are developed aiming to minimize the number of queried weights while providing guarantees on the computed solution. We show in this work the hardness of the underlying problem in general while providing several efficient algorithms that can make use of natural assumptions about the order in which the nodes are processed by the greedy algorithms. Our motivations for exploring this problem stem from finding practical solutions to maximum-weight matching in hypergraphs, a problem recently emerging in the formation of peer-to-peer energy sharing communities.

We study a well-known communication abstraction called Byzantine Reliable Broadcast (BRB). This abstraction is central in the design and implementation of fault-tolerant distributed systems, as many fault-tolerant distributed applications require communication with provable guarantees on message deliveries. Our study focuses on fault-tolerant implementations for message-passing systems that are prone to process-failures, such as crashes and malicious behavior. At PODC 1983, Bracha and Toueg, in short, BT, solved the BRB problem. BT has optimal resilience since it can deal with t

Applications for adaptive (sometimes also called smart) Cyber-Physical Systems are blossoming thanks to the large volumes of data, sensed in a continuous fashion, in large distributed systems. The benefits of these applications come nonetheless with a price: the need for jointly addressing challenges in efficient data communication and analysis (among others). The goal of the DRIVEN framework, presented here, is to address these challenges for a data gathering and distance-based clustering tool in the context of vehicular networks. Because of the limited communication bandwidth (compared to the volume of sensed data) of vehicular networks and the monetary costs of data transmission, the intuition behind DRIVEN is to avoid gathering the data to be clustered in a raw format from each vehicle, but rather to allow for a streaming-based error-bounded approximation, through Piecewise Linear Approximation, to compress the volumes of data to be gathered. At the same time, rather than relying on a batch-based clustering algorithm that requires all the data to be first gathered (and then clustered), DRIVEN relies on and extends a streaming-based clustering algorithm that leverages the inherent ordering of the spatial and temporal data being collected, to perform the clustering in an online fashion, while data is being retrieved. As we show, based on our prototype implementation using Apache Flink and our evaluation with real-world data such as GPS and LiDAR, the accuracy loss for the clustering performed on the reconstructed data can be small, even when the raw data is compressed to 10-35% of its original size, and the transferring of data itself can be completed in up to one-tenth of the duration observed when gathering raw data.