1,720,980 research outputs found
Massive Opportunistic Sensing with Limited Collaboration for Age of Information
No full textWe consider an Internet of thing scenario, where a set of sensors collect data and exchange them with a common receiver. We analyze their interaction, considering a shared goal to minimize Age of information at the receiver's side. We argue that a fully collaborative setup, albeit generally succeeding in this task at first, often leads to resource wastage in the long run. We try to achieve a similar level of cooperation through a purely opportunistic mechanism, in which nodes are driven by selfish objectives, but still aware of the ultimate goal of maximizing information freshness. We show how our proposed approach, allowing fewer nodes to participate in the task (up to one order of magnitude), results in a better resource management, still improving the long-term average age of information. At the same time, a target number of participating nodes can be set, e.g., to a given fraction of the network, by properly tuning the individual objectives and the communication costs
Analysis of Age of Information in Slotted ALOHA Networks With Different Strategic Backoff Schemes
No full textStatus update freshness in slotted ALOHA networks is an important issue for Internet of things scenarios with large number of nodes and uncoordinated access. We compare the age of information of three different implementations of a backoff to counteract collisions due to uncoordinated medium access, where the transmission probability is (i) gradually decreased, (ii) turned to 0 after a collision, or (iii) turned to 0 proactively. We discuss whether these strategies decrease the average AoI of the nodes, and highlight how their efficiency changes with a distributed application in a game theoretic fashion. As a result, the gradual backoff scheme is not recommended, whereas the reactive scheme has an optimal performance inferior to the proactive one, but obtains analogous results at the Nash equilibrium, and can be a candidate for practical implementations
Cybersecurity Analysis Through Shapley Values for a Network Traffic Dataset of Android Malware
No full textWe explore the use of machine learning, specifically Random Forest classifiers, combined with SHapley Additive exPlanations values, to detect Android malware. We leverage diverse datasets, including the Android Genome Project and Drebin, to distinguish between benign and malicious applications. Emphasizing feature importance through SHAP analysis, we aim to enhance model interpretability and effectiveness in cybersecurity. This approach not only improves threat detection accuracy, but also contributes to the broader field of explainable AI in cybersecurity. The paper is structured to cover theoretical foundations, methodology, results, and future directions in this evolving area of study. Also, based on practical findings, we highlight the importance of the data source and transmission patterns as a way to identify malware
Machine Learning Misclassification Within Status Update Optimization
No full textThis paper explores the optimization of status updates in sensing systems, focusing on misclassification in machine learning (ML) models. Previous research has primarily tackled the impact of different techniques throughout the communication layers on Age of Information (AoI), or alternatively studied the Age of Incorrect Information (AoII) as a flaw that can be counteracted by a more active transmission pattern. Our study presents analytical considerations, as well as simulation results from real datasets, with the original aspect that classification errors are not an externality, but are triggered by a fraction of the status updates themselves, which therefore ought to be kept under control. An excessively high number of transmission may be damaging the system, and the right balance needs to be found between prompt updating that lowers AoI, and accuracy to minimize AoII at the same time. In this sense, we offer a new standpoint for timely status update, where freshness of correct information is required for smart systems to make the best decision in real-time
DCP: a TCP-Inspired Method for Online Domain Adaptation under Dynamic Data Drift
No full textMobile computing faces challenges due to the resource constraints of mobile devices, such as limited computing power, energy, and connectivity. These limitations hinder the use of high-complexity classifiers and wireless transmissions. To address this issue, we propose a novel collaboration paradigm between mobile devices and edge servers, where the edge server assists the mobile devices by dynamically retraining a low-complexity classifier to adapt to temporal changes in data distribution. We propose a novel approach called drift control protocol (DCP) which is inspired by TCP congestion control mechanism. DCP aims to strike a balance between low-complexity classifier retraining frequency and communication costs with the edge server. It adjusts the update rate of the classifier on the mobile device based on distribution drift characteristics and controls the number of input samples sent to the edge server to improve accuracy. We evaluate and study different versions of DCP using synthetic and real datasets We demonstrate that DCP keeps the error bound, while reducing the burden of the communication cost by 90% for the mobile nodes, which makes our proposal suitable for online domain adaptation
Strategic Backoff of Slotted ALOHA for Minimal Age of Information
Full text linkRandom access protocols are usually adopted in the Internet of Things to enable uncoordinated medium sharing. Tackling this setting, we explore the statistics of the packet inter-delivery times under slotted ALOHA contention, considering two backoff schemes (reactive vs. proactive). We further discuss the efficiency of these schemes in minimizing the average age of information. Finally, we investigate age minimization both as a centralized optimization and via game theory, obtaining numerical solutions for both cases. A reactive scheme applied in a centralized manner is found to be the most suitable to systems that require a bounded age, whereas a proactive solution applied distributedly is best used to minimize the average age
Mitigating the feature fatigue effect for smart products through digital servitization
No full textWe analyze the problem of a manufacturer developing smart products and selling them to consumers who can be subject to the feature fatigue effect. Feature fatigue occurs with the presence of many features (smart products are typically over featured goods) that make consumers excited during the purchasing phase while becoming useless and complicated during the consumption phase. The latter negative implications induce frustration and regret among consumers, who can associate these negative feelings and effects with the brand, leading to a detrimental impact on its value. We capture the effect of feature fatigue on brand value through a dynamic model in which the stock of goodwill decreases with excessive investments in technology, which are the main drivers to create fashionable, attractive, and appealing features. We propose a digital servitization strategy to mitigate these negative effects and discover that firms offering digital servitization can end up economically worse off because feature fatigue plays a key role in the optimal strategies. In contrast, firms are economically better off and resolve feature fatigue when digital servitization is independent of other strategies. We identify cases in which digital servitization does not provide any benefit, when it completely nullifies feature fatigue effect, and when its activation complements either short-term or long-term strategies. Our findings are then supported and proofed empirically to increase their generizability
Joint Communication and Inference User Allocation in LLM Native Networks
No full textLarge language models (LLMs) are game changers for future next-generation networks, unlocking new opportunities for disruptive and interactive services and applications. Edge computing enables deployment of LLMs closer to the users, allowing for the implementation of highly responsive intelligent systems. This paper proposes a matching theory-based algorithm to optimize the user-LLM association and considers both the communication and inference delay, in the presence of capacity-constrained edge nodes. The objective is to minimize end-to-end user delay, that is, the time elapsed between when a user submits a request and when the response is sent back. Therefore, a matching game is formulated between the users and the LLMs, assuming heterogeneous LLMs, specialized in different types of learning tasks. The scenario is modeled as a matching game with externalities and incomplete lists, which terminates in a stable configuration, leveraging monotonic user preference list metric, within the algorithm execution. A comparative performance evaluation against different state-of-the-art techniques confirms the advantages of adopting a joint communication and inference aware approach to orchestrate the user-LLM assignments
- …
