Dynamic Thermal Management (DTM) systems are critical to the reliable operation of Multiprocessor System-on-Chips (MPSoCs), yet remain vulnerable to sophisticated thermal manipulation attacks. These attacks, executed through hardware trojans or privilege escalation, can compromise the integrity of thermal sensors, causing performance degradation, accelerated aging, and catastrophic hardware failure by disabling thermal throttling mechanisms. Existing countermeasures rely on reactive detection methods and conventional machine learning models that fail to capture the complex physics governing thermal systems, including thermal coupling across cores and power-frequency interdependencies, making them ineffective against multi-stage attacks that exploit DTM decision-making logic. This work presents a novel transformer-based defense framework that leverages self-attention mechanisms to model rich, system-wide feature interactions for detecting adversarial thermal manipulations in real time. The proposed hybrid architecture integrates an adaptive pre-filtering with dynamic thresholding to achieve an 83x throughput improvement (22,798 samples/second versus 274.53 samples/second for transformer-only baseline) and nearly 50% lower GPU utilization, enabling deployment on resource-constrained embedded platforms. Comprehensive on-device validation on the NVIDIA Jetson AGX Orin board demonstrates substantial thermal regulation improvements, reducing average peak temperatures from 103 °C to 98.5 °C while maintaining a model active ratio of only 2.73%. The framework incorporates an adaptive defense system with load-dependent dynamic thresholding that achieves high F1-scores in detecting the thermal attacks discussed in the literature (0.75 to 0.9). This work bridges the critical gap between simulation-based security research and practical embedded system deployment, establishing a new paradigm for lightweight, attention-based anomaly detection in thermally constrained environments.

Efficient thermal and power management in modern multiprocessor systems-on-chip (MPSoCs) demands accurate power consumption estimation. One of the state-of-theart approaches, Alternative Blind Power Identification (ABPI), theoretically eliminates the dependence on steady-state temperatures, addressing a major shortcoming of previous approaches. However, ABPI performance has remained unverified in actual hardware implementations. In this study, we conducted the first empirical validation of ABPI on commercial hardware using the NVIDIA Jetson Xavier AGX platform. Our findings reveal that, while ABPI provides computational efficiency and independence from steady-state temperature, it exhibits considerable accuracy deficiencies in real-world scenarios. To overcome these limitations, we introduce a novel approach that integrates Custom Physics-Informed Neural Networks (CPINNs) with the underlying thermal model of ABPI. Our approach utilizes a specialized loss function that constrains the data-driven model according to the system's thermal physics, complemented by NSGA-II multi-objective optimization with efficient convergence strategies to balance estimation accuracy and computational cost. In experimental validation, CPINN-ABPI achieves a reduction of 84.7% CPU and 73.9% GPU in the mean absolute error (MAE) relative to ABPI, with the weighted mean absolute percentage error (WMAPE) improving from 47%-81% to ~ 12%. The method maintains real-time performance with 195.3µs of inference time. Extensive evaluation on unseen benchmarks demonstrates robust generalization without overfitting.

Modern multicore System-on-Chips (SoCs) include hardware monitoring mechanisms to measure total power consumption, but these aggregate measurements are insufficient for fine-grained thermal and power management. This paper introduces an improved Clustering Blind Power Identification (ICBPI), an approach to improve the sensitivity and robustness of the Blind Power Identification (BPI) approach, which identifies the power consumption of different cores and the thermal model of an SoC using only thermal sensor measurements and the total power consumption. The proposed approach enhances BPI’s initialization step (specifically the non-negative matrix factorization, which is crucial for BPI accuracy) by incorporating density-based spatial clustering of of noise applications (DBSCAN). This is done to maximize the physical relationship between the temperature and power consumption, ensuring more accurate power estimates. Our simulations demonstrate two tasks to validate the proposed approach. The first evaluates the power accuracy per core on four different multicores, including a heterogeneous processor, showing that ICBPI significantly improves accuracy without overheads. For example, in a four-core SoC, error rates are reduced by 77.56% compared to vanilla BPI and by 68.44% compared to the state-of-the-art approach called BPISS. The second task focuses on enhancing the precision and robustness of the detection and localization of malicious thermal sensor attacks in the heterogeneous processor, demonstrating that ICBPI is capable of enhancing security of multicore SoCs.

In the task of modeling user preferences for movie recommender systems, recent research has demonstrated the benefits of describing movies with their eudaimonic and hedonic scores (E and H scores), which reflect the depth of their message and the level of fun experience they provide, respectively. So far, the labeling of movies with their E and H scores has been done manually using a dedicated instrument (a questionnaire), which is time-consuming. To address this issue, we propose an automatic approach for predicting E and H scores. Specifically, we collected E and H scores of 709 movies from 370 users (with a total of 3699 records), augmented this dataset with metadata, audio, and low-level and high-level visual features, and trained machine learning models for predicting the E and H scores of movies. This study investigates the use of machine learning models in predicting the E and H scores of movies using various feature sets, including audio, low-level and high-level visual features, and metadata. We compared the performance of predictive models using different combinations of features with the majority classifier as the baseline approach. The results demonstrate that our proposed machine learning-based models significantly outperform the baseline in predicting E and H scores, particularly when leveraging metadata features. Specifically, the random forest classifier achieved a 20% increase in ROC AUC compared to the baseline when predicting both the E score and the H score. These improvements were found to be statistically significant. Overall, our findings suggest that automated tools for predicting E and H scores in movies are promising alternatives to traditional questionnaire-based approaches.

The structural change toward the digital transformation of online sales elevates the importance of parallel processing techniques in recommender systems, particularly in the pandemic and post-pandemic era. Matrix factorization (MF) is a popular and scalable approach in collaborative filtering (CF) to predict user preferences in recommender systems. Researchers apply Stochastic Gradient Descent (SGD) as one of the most famous optimization techniques for MF. Paralleling SGD methods help address big data challenges due to the wide range of products and the sparsity in user ratings. However, these methods’ convergence rate and accuracy are affected by the dependency between the user and item latent factors, specifically in large-scale problems. Besides, the performance is sensitive to the applied learning rates. This article proposes a new parallel method to remove dependencies and boost speed-up by using fractional calculus to improve accuracy and convergence rate. We also apply adaptive learning rates to enhance the performance of our proposed method. The proposed method is based on Compute Unified Device Architecture (CUDA) platform. We evaluate the performance of our proposed method using real-world data and compare the results with the close baselines. The results show that our method can obtain high accuracy and convergence rate in addition to high parallelism.

Abstract Recommender systems help people find relevant content in a personalized way. One main promise of such systems is that they are able to increase the visibility of items in the long tail , i.e., the lesser-known items in a catalogue. Existing research, however, suggests that in many situations today’s recommendation algorithms instead exhibit a popularity bias , meaning that they often focus on rather popular items in their recommendations. Such a bias may not only lead to the limited value of the recommendations for consumers and providers in the short run, but it may also cause undesired reinforcement effects over time. In this paper, we discuss the potential reasons for popularity bias and review existing approaches to detect, quantify and mitigate popularity bias in recommender systems. Our survey, therefore, includes both an overview of the computational metrics used in the literature as well as a review of the main technical approaches to reduce the bias. Furthermore, we critically discuss today’s literature, where we observe that the research is almost entirely based on computational experiments and on certain assumptions regarding the practical effects of including long-tail items in the recommendations.

Despite their proven various benefits, Recommender Systems can cause or amplify certain undesired effects. In this paper, we focus on Popularity Bias, i.e., the tendency of a recommender system to utilize the effect of recommending popular items to the user. Prior research has studied the negative impact of this type of bias on individuals and society as a whole and proposed various approaches to mitigate this in various domains. However, almost all works adopted offline methodologies to evaluate the effectiveness of the proposed approaches. Unfortunately, such offline simulations can potentially be rather simplified and unable to capture the full picture. To contribute to this line of research and given a particular lack of knowledge about how debiasing approaches work not only offline, but online as well, we present in this paper the results of user study on a national broadcaster movie streaming platform in Norway, i.e., TV 2, following the A/B testing methodology. We deployed an effective mitigation approach for popularity bias, called Calibrated Popularity (CP), and monitored its performance in comparison to the platform’s existing collaborative filtering recommendation approach as a baseline over a period of almost four months. The results obtained from a large user base interacting in real-time with the recommendations indicate that the evaluated debiasing approach can be effective in addressing popularity bias while still maintaining the level of user interest and engagement.

Hybrid recommender systems utilize advanced algorithms capable of learning heterogeneous sources of data and generating personalized recommendations for users. The data can range from user preferences (e.g., ratings or reviews) to item content (e.g., description or category). Prior studies in the field of recommender systems have primarily relied on ”ratings” as the user feedback, when building user profiles or evaluating the quality of the recommendation. While ratings are informative, they may still fail to represent a comprehensive picture of actual user preferences. In contrast, there are other types of feedback data that differently or complementarily represent users and their preferences, including the reviews and the sentiments encapsulated within them. Such data can reveal important parts of a user’s profile that are not necessarily correlated with user ratings, and hence, they potentially reflect a different side of the user’s profile. In this paper, we propose a novel form of hybrid recommender system, capable of analyzing the reviews and extracting their sentiments that are incorporated into the recommendation process. We used advanced algorithms to generate recommendations for users capable of incorporating additional data, such as the review sentiment. We conducted analyses and showed that sentiments of user reviews are not always highly correlated with the ratings (e.g., in music domain). This might mean that sentiment can be indicative of a different aspect of user preferences and can be used as an alternative signal of user feedback. Hence, we have used both ratings and sentiments of reviews when evaluating our proposed hybrid recommender system. We selected two common datasets for the evaluation, Amazon Digital Music and Amazon Video Games, and showed the superior performance of the proposed hybrid recommender system compared to different baselines. The comparison were made in two evaluation scenarios, namely, when the ratings were considered the user feedback and when sentiments of the review were considered the user feedback.

Recommender Systems are becoming an inherent part of today’s Internet. They can be found anywhere from e-commerce platforms (eBay, Amazon) to music or movie streaming (Spotify, Netflix), social media (Facebook, Instagram, TikTok), travel platforms (Booking.com, Expedia), and much more. Whether a recommendation is successful or not can rely on multiple objectives such as user satisfaction, business value, and societal issues. In addition, the long-term happiness (along with short-term excitements and delight) of the users is critical for a recommender system to be considered successful. MORS workshop brings together researchers and practitioners to discuss the importance of these aspects of recommender systems and find ways to develop algorithms to build multi-objective recommenders and also evaluation metrics to assess their success.

Historically, the main criterion for a successful recommender system was the relevance of the recommended items to the user. In other words, the only objective for the recommendation algorithm was to learn user’s preferences for different items and generate recommendations accordingly. However, real-world recommender systems are well beyond a simple objective and often need to take into account multiple objectives simultaneously. These objectives can be either from the users’ perspective or they could come from other stakeholders such as item providers or any party that could be impacted by the recommendations. Such multi-objective and multi-stakeholder recommenders present unique challenges and these challenges were the focus of the MORS workshop.

Abstract Over the past years, the research of video recommender systems (RSs) has been mainly focussed on the development of novel algorithms. Although beneficial, still any algorithm may fail to recommend video items that the system has no form of data associated to them (New Item Cold Start). This problem occurs when a new item is added to the catalogue of the system and no data are available for that item. In content‐based RSs, the video items are typically represented by semantic attributes, when generating recommendations. These attributes require a group of experts or users for annotation, and still, the generated recommendations might not capture a complete picture of the users' preferences, for example, the visual tastes of users on video style. This article addresses this problem by proposing recommendation based on novel visual features that do not require human annotation and can represent visual aspects of video items. We have designed a novel evaluation methodology considering three realistic scenarios, that is, (a) extreme cold start, (b) moderate cold start and (c) warm‐start scenario. We have conducted a set of comprehensive experiments, and our results have shown the superior performance of recommendations based on visual features, in all of the evaluation scenarios.

Internet of Things (IoT) is composed of physical devices, communication networks, and services provided by edge systems and over-the-top applications. IoT connects billions of devices that collect data from the physical environment, which are pre-processed at the edge and then forwarded to processing services at the core of the infrastructure, on top of which cloud-based applications are built and provided to mobile end users. IoT comes with important advantages in terms of applications and added value for its users, making their world smarter and simpler. These advantages, however, are mitigated by the difficulty of guaranteeing IoT trustworthiness, which is still in its infancy. IoT trustworthiness is a must especially in critical domains (e.g., health, transportation) where humans become new components of an IoT system and their life is put at risk by system malfunctioning or breaches. In this article, we put forward the idea that trust in IoT can be boosted if and only if its automation and adaptation processes are based on trustworthy data. We therefore depart from a scenario that considers the quality of a single decision as the main goal of an IoT system and consider the trustworthiness of collected data as a fundamental requirement at the basis of a trustworthy IoT environment. We therefore define a methodology for data collection that filters untrusted data out according to trust rules evaluating the status of the devices collecting data and the collected data themselves. Our approach is based on blockchain and smart contracts and collects data whose trustworthiness and integrity are proven over time. The methodology balances trustworthiness and privacy and is experimentally evaluated in real-world and simulated scenarios using Hyperledger fabric blockchain.

We present a demo application of a web-based recommender systems that is powered by the “Movie Genome”, i.e., a rich semantic description of a movie's content, including state-of-the-art audio and visual descriptors and metadata (genre and tags). The current version of the Movie Genome web application implements content-based filtering approaches. Due to its modular implementation and the free availability of its source code, it can be easily extended to various context-aware and user-aware scenarios. A personality questionnaire is already integrated into the web application, which allows it to also serve as testbed for personality-aware recommendation algorithms.

Users of a recommender system may be requested to express their preferences about items either with evaluations of items (e.g. a rating) or with comparisons of item pairs. In this work we focus on the acquisition of pairwise preferences in the music domain. Asking the user to explicitly compare music, i.e., which, among two listened tracks, is preferred, requires some user effort. We have therefore developed a novel approach for automatically extracting these preferences from the analysis of the facial expressions of the users while listening to the compared tracks. We have trained a predictor that infers user's pairwise preferences by using features extracted from these data. We show that the predictor performs better than a commonly used baseline, which leverages the user's listening duration of the tracks to infer pairwise preferences. Furthermore, we show that there are differences in the accuracy of the proposed method between users with different personalities and we have therefore adapted the trained model accordingly. Our work shows that by introducing a low user effort preference elicitation approach, which, however, requires to access information that may raise potential privacy issues (face expression), one can obtain good prediction accuracy of pairwise music preferences.

In this paper, we explore the potential of using visual features in movie Recommender Systems. This type of content features can be extracted automatically without any human involvement and have been shown to be very effective in representing the visual content of movies. We have performed the following experiments, using a large dataset of movie trailers: (i) Experiment A: an exploratory analysis as an initial investigation on the data, and (ii) Experiment B: building a movie recommender based on the visual features and evaluating the performance. The observed results have shown promising potential of visual features in representing the movies and the excellency of recommendation based on these features.

The ACM Recommender Systems Challenge 20171 focused on the problem of job recommendations: given a new job advertisement, the goal was to identify those users who are both (a) interested in getting notified about the job advertisement, and (b) appropriate candidates for the given job. Participating teams had to balance between user interests and requirements for the given job as well as dealing with the cold-start situation. For the first time in the history of the conference, the RecSys challenge offered an online evaluation: teams first had to compete as part of a traditional offline evaluation and the top 25 teams were then invited to evaluate their algorithms in an online setting, where they could submit recommendations to real users. Overall, 262 teams registered for the challenge, 103 teams actively participated and submitted together more than 6100 solutions as part of the offline evaluation. Finally, 18 teams participated and rolled out recommendations to more than 900,000 users on XING2.

In the last years, there has been much attention given to the semantic gap problem in multimedia retrieval systems. Much effort has been devoted to bridge this gap by building tools for the extraction of high-level, semantics-based features from multimedia content, as low-level features are not considered useful because they deal primarily with representing the perceived content rather than the semantics of it.

We present a research tool for user preference elicitation that collects both explicit user feedback and unobtrusively acquired facial expressions. The concrete implementation is a web-based user interface where the user is presented with two music excerpts. After listening to both, the user provides a pairwise score (i.e. which of the two items is preferred) for each pair of music excerpts. The novelty of the demo is the integration of the unobtrusive acquisition of facial expressions through the webcam. During the listening of the music excerpts, the system extracts features related to the facial expressions of the user several times per second. The interaction runs as a web application, which allows for a large-scale remote acquisition of emotional data. Up to now, such acquisitions were usually done in controlled environments with few subjects, hence being of little use for the recommender systems community.

Traditional food recommender systems exploit items' ratings and descriptions in order to generate relevant recommendations for the users. While this data is important, it might not entirely capture the true users' preferences. In this paper, we analyse the performance of a food recommender that allows users to enter their preferences in the form of both ratings and tags, which are then used by a Matrix Factorization (MF) rating prediction model. The performed offline and online experiments have clarified the importance of user tags in comparison to content features. While item content contributes more to the quality of the prediction accuracy, user tags yields better ranking quality. Even more importantly, a live user study has revealed that a system variant, which leverages user tags in the prediction model and in the interface, achieves a significantly better user evaluation in terms of perceived effectiveness, choice satisfaction and choice difficulty.

The new user problem in recommender systems is still challenging, and there is not yet a unique solution that can be applied in any domain or situation. In this paper we analyze viable solutions to the new user problem in collaborative filtering (CF) that are based on the exploitation of user personality information: (a) personality-based CF , which directly improves the recommendation prediction model by incorporating user personality information, (b) personality-based active learning , which utilizes personality information for identifying additional useful preference data in the target recommendation domain to be elicited from the user, and (c) personality-based cross-domain recommendation , which exploits personality information to better use user preference data from auxiliary domains which can be used to compensate the lack of user preference data in the target domain. We benchmark the effectiveness of these methods on large datasets that span several domains, namely movies, music and books. Our results show that personality-aware methods achieve performance improvements that range from 6 to 94 % for users completely new to the system, while increasing the novelty of the recommended items by 3–40 % with respect to the non-personalized popularity baseline. We also discuss the limitations of our approach and the situations in which the proposed methods can be better applied, hence providing guidelines for researchers and practitioners in the field.

The accuracy of collaborative-filtering recommender systems largely depends on three factors: the quality of the rating prediction algorithm, and the quantity and quality of available ratings. While research in the field of recommender systems often concentrates on improving prediction algorithms, even the best algorithms will fail if they are fed poor-quality data during training, that is, garbage in, garbage out. Active learning aims to remedy this problem by focusing on obtaining better-quality data that more aptly reflects a user's preferences. However, traditional evaluation of active learning strategies has two major flaws, which have significant negative ramifications on accurately evaluating the system's performance (prediction error, precision, and quantity of elicited ratings). (1) Performance has been evaluated for each user independently (ignoring system-wide improvements). (2) Active learning strategies have been evaluated in isolation from unsolicited user ratings (natural acquisition). In this article we show that an elicited rating has effects across the system, so a typical user-centric evaluation which ignores any changes of rating prediction of other users also ignores these cumulative effects, which may be more influential on the performance of the system as a whole (system centric). We propose a new evaluation methodology and use it to evaluate some novel and state-of-the-art rating elicitation strategies. We found that the system-wide effectiveness of a rating elicitation strategy depends on the stage of the rating elicitation process, and on the evaluation measures (MAE, NDCG, and Precision). In particular, we show that using some common user-centric strategies may actually degrade the overall performance of a system. Finally, we show that the performance of many common active learning strategies changes significantly when evaluated concurrently with the natural acquisition of ratings in recommender systems.

This demo paper presents a context-aware recommendation system. The system mines data from user's web searches and other sources to improve the presentation of content on visited web pages. While user is browsing the internet, a memory resident agent records and analyzes the content of the webpages that were either searched for or visited in order to identify topic preferences. Then, based on such information, the content of requested web page is ranked and classified with different styles. The demo shows how a music weblog can be modified automatically based on user's affinities.

We demonstrate a system for filtering media streams according to the collective taste of a leader-less informal clan of users. Applications on mobile devices receive streams of content items that are assessed by local software agents. The agents learn the collective preferences of the tribe by forming a distributed multi-agent society that shares data on the behavior of all users. The underlying artificial intelligence is based on support vector machines that cooperate by broadcasting new support vectors. The demo shows micro-blog readers on cell phones running support vector machine agents with text kernels and communicating over IP Multimedia System networks.