The use of DNS over HTTPS (DoH) tunneling by an attacker to hide malicious activity within encrypted DNS traffic poses a serious threat to network security, as it allows malicious actors to bypass traditional monitoring and intrusion detection systems while evading detection by conventional traffic analysis techniques. Machine Learning (ML) techniques can be used to detect DoH tunnels; however, their effectiveness relies on large datasets containing both benign and malicious traffic. Sharing such datasets across entities is challenging due to privacy concerns. In this work, we propose CO-DEFEND (Continuous Decentralized Federated Learning for Secure DoH-Based Threat Detection), a Decentralized Federated Learning (DFL) framework that enables multiple entities to collaboratively train a classification machine learning model for DoH threat detection while preserving data privacy, enhancing scalability and resilience against single points of failure. The proposed DFL framework provides a realistic implementation for DoH threat detection, enabling multiple entities to train their local models online with incoming DoH flows in real-time batches as they are processed – an approach that fits naturally within modern Internet architectures. This framework adapts four classical machine learning algorithms, Support Vector Machines (SVM), Logistic Regression (LR), Decision Trees (DT), and Random Forest (RF), for federated scenarios and efficient training. In addition, a key methodological feature of CO-DEFEND is the use of DT and RF as model selection rather than aggregation mechanisms, allowing each participant to retain interpretable and locally optimal decision structures while benefiting from collective updates. We compare our proposed method by using the dataset CIRA-CIC-DoHBrw-2020 with existing machine learning approaches, including more computationally complex alternatives such as neural networks, to demonstrate its effectiveness in detecting malicious DoH tunnels while improving scalability and computational efficiency.

Distributed AI and IoT applications increasingly execute across heterogeneous resources spanning end devices, edge/fog infrastructure, and cloud platforms, often under different administrative domains. Fluid Computing has emerged as a promising paradigm for enhancing massive resource management across the computing continuum by treating such resources as a unified fabric, enabling optimal service-agnostic deployments driven by application requirements. However, existing solutions remain largely centralized and often do not explicitly address multi-domain considerations. This paper proposes an agnostic multi-domain orchestration architecture for fluid computing environments. The orchestration plane enables decentralized coordination among domains that maintain local autonomy while jointly realizing intent-based deployment requests from tenants, ensuring end-to-end placement and execution. To this end, the architecture elevates domain-side control services as first-class capabilities to support application-level enhancement at runtime. As a representative proof of concept, we instantiate the architecture through a distributed AI use case; specifically, we consider a multi-domain Decentralized Federated Learning (DFL) deployment under Byzantine threats. Under this setting, we leverage domain-side capabilities to enhance Byzantine security by introducing FU-HST, an SDN-enabled multi-domain anomaly detection mechanism that complements Byzantine-robust aggregation. We validate the use-case workflow via simulation in single- and multi-domain settings, evaluating anomaly detection, DFL performance, and computation/communication overhead.

The Messaging Layer security (MLS) and its underlying Continuous Group Key Agreement (CGKA) protocol allows a group of users to share a cryptographic secret in a dynamic manner, such that the secret is modified in member insertions and deletions. Although this flexibility makes MLS ideal for implementations in distributed environments, a number of issues need to be overcome. Particularly, the use of digital certificates for authentication in a group goes against the group members' privacy. In this work we provide an alternative method of authentication in which the solicitors, instead of revealing their identity, only need to prove possession of certain attributes, dynamically defined by the group, to become a member. Instead of digital certificates, we employ Attribute-Based Credentials accompanied with Selective Disclosure in order to reveal the minimum required amount of information and to prevent attackers from linking the activity of a user through multiple groups. We formally define a CGKA variant named Attribute-Authenticated Continuous Group Key Agreement (AA-CGKA) and provide security proofs for its properties of Requirement Integrity, Unforgeability and Unlinkability. We also provide an implementation of our AA-CGKA scheme and show that it achieves performance similar to a trivial certificate-based solution.

The placement of Cloud-Native Network Functions (CNFs) across the Cloud-Continuum represents a core challenge in the orchestration of current 5G and future 6G networks. The process entails the implementation of interdependent computing tasks, which are structured as Service Function Chains, over distributed cloud infrastructures. This is achieved while satisfying strict resource, bandwidth, connectivity, and end-to-end latency constraints. It is widely acknowledged that classical approaches, including mixed-integer (non)linear programming (MINLP), heuristics, and reinforcement learning, face practical limitations in terms of scalability, robust constraint handling, and generalization to unseen network conditions. In this study, a diffusion-based theoretical and algorithmic framework for CNF placement is proposed, based on Denoising Diffusion Probabilistic Models (DDPM). The placement process is conceptualized as a conditional graph-to-assignment generation task. Each scenario is encoded as a heterogeneous graph, capturing infrastructure and service-chain structure. A Graph Neural Network (GNN) denoiser is trained to iteratively refine noisy CNF-to-cloud assignment matrices. In order to bias the generation process towards valid deployments, the model incorporates constraint-aware penalties during training. At inference, a multitude of candidate placements are sampled, and the best suboptimal, feasible solution is selected. This enables a controllable trade-off between solution quality and runtime. Extensive experimentation on diverse topologies, incorporating out-of-distribution evaluations with larger instances and shifted constraint regimes, demonstrates that the proposed approach consistently generates feasible solutions with considerably accelerated inference compared to MINLP solvers when available, while maintaining robust feasibility under constraint-tight scenarios. The findings of this study demonstrate the potential of diffusion-based generative modelling as a scalable tool for constrained network placement and embedding in cloud-continuum orchestration.

The advance of quantum computing poses a significant threat to classical cryptography, compromising the security of current encryption schemes such as RSA and ECC. In response to this challenge, two main approaches have emerged: quantum cryptography and post-quantum cryptography (PQC). However, both have implementation and security limitations. In this paper, we propose a secure key distribution protocol for Quantum Key Distribution Networks (QKDN), which incorporates encapsulation techniques in the key-relay model for QKDN inspired by onion routing and combined with PQC to guarantee confidentiality, integrity, authenticity and anonymity in communication. The proposed protocol optimizes security by using post-quantum public key encryption to protect the shared secrets from intermediate nodes in the QKDN, thereby reducing the risk of attacks by malicious intermediaries. Finally, relevant use cases are presented, such as critical infrastructure networks, interconnection of data centers and digital money, demonstrating the applicability of the proposal in critical highsecurity environments.

Fluid Computing has emerged as a promising paradigm for enhancing massive and heterogeneous resource management across the Cloud-to-Edge continuum for Internet of Things (IoT) and artificial intelligence (AI) applications. Despite its advantages, research into the optimal deployment of distributed applications across fluid scenarios remains scarce, and existing centralized frameworks cannot exploit emerging AI-native features nor model realistic multi-domain deployment scenarios. This paper presents an innovative provider-based architecture for the optimal orchestration of distributed AI services in fluid environments under 6G network capabilities. The proposed hybrid solution includes robust and scalable decentralized orchestration for the placement of cross-provider workloads without a central broker, as well as leveraging autonomous decision-making devices through distributed task offloading techniques. The proposal was tailored to a Decentralized Federated Learning deployment, serving as a use case in settings with strict privacy and security requirements. This approach was adopted to illustrate the viability of the proposal by deploying large distributed AI services in 6G-like networks.

The placement of Cloud-Native Network Functions (CNFs) across the Cloud-Continuum represents a core challenge in the orchestration of current 5G and future 6G networks. The process involves the placement of interdependent computing tasks, structured as Service Function Chains, over distributed cloud infrastructures. This is achieved while satisfying strict resource, bandwidth and latency constraints. It is acknowledged that classical approaches, including mixed-integer nonlinear programming, heuristics and reinforcement learning are limited in terms of scalability, constraint handling and generalisation capacity. In the present study, a novel theoretical framework is proposed, which is based on Denoising Diffusion Probabilistic Models (DDPM) for CNF placement. The present approach proposes a reconceptualisation of the placement problem as a generative graph to assignment task, where the placement problem is encoded as a heterogeneous graph, and a Graph Neural Network denoiser is trained to iteratively refine noisy CNF-to-cloud assignment matrices. The model incorporates constraint-specific losses directly into the loss function, thereby allowing it to learn feasible solution spaces. The integration of the DDPM formulation with structured combinatorial constraints is achieved through a rigorous and systematic approach. Extensive evaluations across diverse topologies have been conducted, which have confirmed that the model consistently produces feasible solutions with orders of magnitude faster inference than MINLP solvers. The results obtained demonstrate the potential of diffusion-based generative modelling for constrained network embedding problems, making an impact towards the practical, scalable orchestration of distributed Cloud-Native Network Functions.

The rapid growth of Internet of Things (IoT) devices and applications has led to an increased demand for advanced analytics and machine learning techniques capable of handling the challenges associated with data privacy, security, and scalability. Federated learning (FL) and blockchain technologies have emerged as promising approaches to address these challenges by enabling decentralized, secure, and privacy-preserving model training on distributed data sources. In this paper, we present a novel IoT solution that combines the incremental learning vector quantization algorithm (XuILVQ) with Ethereum blockchain technology to facilitate secure and efficient data sharing, model training, and prototype storage in a distributed environment. Our proposed architecture addresses the shortcomings of existing blockchain-based FL solutions by reducing computational and communication overheads while maintaining data privacy and security. We assess the performance of our system through a series of experiments, showing its potential to enhance the accuracy and efficiency of machine learning tasks in IoT settings.

Federated Learning (FL) is a machine learning framework that enhances privacy compared to normal FL, since model training is done across multiple nodes by exchanging only local parameters instead of raw data. Despite this advantage, FL is still vulnerable to some privacy attacks that have been addressed with techniques such as Differential Privacy (DP), Homomorphic Encryption (HE), or Secure Multi-Party Computation (SMPC). These methods need further assumptions that narrow their range of application, exhibiting problems to work with non-linear functions, to perform large matrix multiplications and having high communication and computational costs to manage semi-honest nodes. In this paper, we propose a solution to guarantee privacy in FL schemes that simultaneously solves these shortcomings. Our proposal is based on the Berrut Approximated Coded Computing (BACC), a technique from the coded computing paradigm, adapted here to provide input privacy to FL in a scalable way. In our Private BACC, controlled randomness is introduced into the rational Berrut approximation of a target function so that the input values are hidden and the output function values can still be computed with good approximation. We provide a theoretical bound to the privacy leakage level of private BACC, and further describe its application for computing arbitrary non-linear functions, particularly distributed matrix multiplications. Our numerical results demonstrate that adding privacy in this form outperforms DP in model quality, since it has negligible cost in the precision. Additionally, it beats HE in terms of computational cost, and requires lower computation and communications costs than with SMPC algorithms. Finally, the proposed method supports distributed matrix multiplications, so it can be integrated into a wide class of machine learning systems.

The advent of 5th Generation (5G) networks has introduced the strategy of network slicing as a paradigm shift, enabling the provision of services with distinct Quality of Service (QoS) requirements. The 5th Generation New Radio (5G NR) standard complies with the use cases Enhanced Mobile Broadband (eMBB), Ultra-Reliable Low Latency Communications (URLLC), and Massive Machine Type Communications (mMTC), which demand a dynamic adaptation of network slicing to meet the diverse traffic needs. This dynamic adaptation presents both a critical challenge and a significant opportunity to improve 5G network efficiency. This paper proposes a Deep Reinforcement Learning (DRL) agent that performs dynamic resource allocation in 5G wireless network slicing according to traffic requirements of the 5G use cases within two scenarios: eMBB with URLLC and eMBB with mMTC. The DRL agent evaluates the performance of different decoding schemes such as Orthogonal Multiple Access (OMA), Non-Orthogonal Multiple Access (NOMA), and Rate Splitting Multiple Access (RSMA) and applies the best decoding scheme in these scenarios under different network conditions. The DRL agent has been tested to maximize the sum rate in scenario eMBB with URLLC and to maximize the number of successfully decoded devices in scenario eMBB with mMTC, both with different combinations of number of devices, power gains and number of allocated frequencies. The results show that the DRL agent dynamically chooses the best decoding scheme and presents an efficiency in maximizing the sum rate and the decoded devices between 84% and 100% for both scenarios evaluated. • 5G Network Slicing Optimization. • eMBB, URLLC, and mMTC Coexistence. • OMA, NOMA, RSMA. • Deep Reinforcement Learning. • The maximization of the sum rate on the scenario eMBB with URLLC and the maximization of the number of successfully decoded devices on the scenario eMBB with mMTC.

High-entropy random numbers are an essential part of cryptography, and Quantum Random Number Generators (QRNG) are an emergent technology that can provide high-quality keys for cryptographic algorithms but unfortunately are currently difficult to access. Existing Entropy-as-a-Service solutions require users to trust the central authority distributing the key material, which is not desirable in a high-privacy environment. In this paper, we present a novel key transmission protocol that allows users to obtain cryptographic material generated by a QRNG in such a way that the server is unable to identify which user is receiving each key. This is achieved with the inclusion of Zero Knowledge Succinct Non-interactive Arguments of Knowledge (zk-SNARK), a cryptographic primitive that allow users to prove knowledge of some value without needing to reveal it. The security analysis of the protocol proves that it satisfies the properties of Anonymity, Unforgeability and Confidentiality, as defined in this document. We also provide an implementation of the protocol demonstrating its functionality and performance, using NFC as the transmission channel for the QRNG key.

Decentralised machine learning has recently been proposed as a potential solution to the security issues of the canonical federated learning approach. In this paper, we propose a decentralised and collaborative machine learning framework specially oriented to resource-constrained devices, usual in IoT deployments. With this aim we propose the following construction blocks. First, an incremental learning algorithm based on prototypes that was specifically implemented to work in low-performance computing elements. Second, two random-based protocols to exchange the local models among the computing elements in the network. Finally, two algorithmics approaches for prediction and prototype creation. This proposal was compared to a typical centralized incremental learning approach in terms of accuracy, training time and robustness with very promising results.

In this paper, we address the optimization of the energy efficiency of underlay multicast device-to-device (D2MD) communications on cellular networks. In particular, we maximize the energy efficiency of both the global network and the individual users considering various fairness factors such as maximum power and minimum rate constraints. For this, we employ a canonical mixed-integer non-linear formulation of the joint power control and resource allocation problem. To cope with its NP-hard nature, we propose a two-stage semi-distributed solution. In the first stage, we find a stable, yet sub-optimal, channel allocation for D2MD groups using a cooperative coalitional game framework that allows co-channel transmission over a set of shared resource blocks and/or transmission over several different channels per D2MD group. In the second stage, a central entity determines the optimal transmission power for each user in the system via fractional programming. We performed extensive simulations to analyze the resulting energy efficiency and attainable transmission rates. The results show that the performance of our semi-distributed approach is very close to that obtained with a pure optimal centralized one.

Social networks analysis allows to study and understand the structural properties of a wide spectrum of natural or artificial systems. In the field of education, online social networks arise quite naturally in the virtual classrooms as an inherent part of the learning activities. In this work we focus in forums participation, modeling and investigating the social relationships taking place during an undergraduate course on computer networks. Our findings show significant correlations among the patterns of engagement and the structure of the networks and the students’ achievements.

In this paper, we optimize the energy efficiency (bits/s/Hz/J) of device-to-multi-device (D2MD) wireless communications. While the device-to-device scenario has been extensively studied to improve the spectral efficiency in cellular networks, the use of multicast communications opens the possibility of reusing the spectrum resources also inside the groups. The optimization problem is formulated as a mixed integer non-linear joint optimization for the power control and allocation of resource blocks (RBs) to each group. Our model explicitly considers resource sharing by letting co-channel transmission over a RB (up to a maximum of r transmitters) and/or transmission through s different channels in each group. We use an iterative decomposition approach, using first matching theory to find a stable even if sub-optimal channel allocation, to then optimize the transmission power vectors in each group via fractional programming. In addition, within this framework, both the network energy efficiency and the max-min individual energy efficiency are investigated. We characterize numerically the energy-efficient capacity region, and our results show that the normalized energy efficiency is nearly optimal (above 90% of the network capacity) for a wide range of minimum-rate constraints. This performance is better than that of other matching-based techniques previously proposed.

In underlay device-to-device multicast communication (D2MD), a group of users can communicate directly by reusing cellular resources blocks (RB) to share common content. This paradigm brings great benefits to cellular networks in term of energy and spectral efficiency. However, D2MD or cellular communication quality may degrade or blocked due to harmful mutual interference between cellular, D2MD users sharing the same communication resources. Therefore, to fully achieve the advantages of D2MD communication, resource management and power control became critical. In this paper, we model the join power and resource allocation as a mixed integer nonlinear problem (MINLP) to maximize network global energy efficiency (GEE). To coup with NP-hard nature of the problem, we investigate a two stages solution. At first, we propose a scheme based on matching theory to solve resource allocation sub-problem. Here, we introduce a reuse and a split factor to control the aggregated interference on CUE from D2MD and the number of RB used by each group. Second, we apply the framework of fractional programming to optimally solve the power control sub-problem subject to QoS constraints. Finally, GEE metric is analysed via extensive numerical simulations with a spatial Poisson process for the users' locations and applying different clustering algorithms as K-nearest neighbour, distance limit.

In this paper, we address the problem of joint resource and power allocation in underlay device-to-device multi-cast communication for short (D2MD) to maximize system global energy efficiency (GEE). For this, we propose a two-stage semi-distributed solution. First, we model the resource sharing sub-problem as a transferable overlapping coalition formation game. Here, a D2MD group can participate in s coalitions and it can decide to join or leave a coalition based on specific split and merge rules. Similarly, a resource block can be shared among r D2MD groups where s, r are reuse and split factors. After that, the transmission power is centrally controlled by the The base station (BS) via a fractional programming framework. Finally, GEE is analysed via extensive numerical simulations with a spatial Poisson process for the users' locations and applying two different clustering algorithms: K-nearest neighbour, and distance-limited.

The IEEE 802.3bj standard defines two potential low power operating modes for high speed energy efficient ethernet (EEE) physical interfaces working at 40 and 100 Gb/s: a not-so-efficient low power mode that requires very short transition times to restore normal operation (Fast-Wake) and a highly efficient low power mode with longer transition times (Deep-Sleep). In this paper, we present a new frame coalescing mechanism that dynamically adjusts the coalescing queue threshold in order to minimize the energy consumption of dual-mode EEE interfaces and maintains, at the same time, the average frame delay close to a target value. The proposed mechanism has been validated through simulation under different types of traffic (Poisson, self-similar, and real Internet traffic). In addition, we show that, with the current transition times and efficiency profiles of the standardized low power modes, our proposal renders the Fast-Wake mode unnecessary in most practical scenarios.

The widespread use of computing and communications technologies has enabled the popularity of social networks oriented to learn. In this work, we study the nature and strength of associations between students using an online social network embedded in a learning management system. With datasets from two offerings of the same course, we mined the sequences of questions and answers posted by the students to identify structural properties of the social graph, patterns of collaboration among students and factors influencing the final achievements. The results show some hints to pursue and investigate deep user analytics in online social learning systems, e.g., to build accurate prediction models for the success of effectiveness of the learning tasks based on the patterns of students' participation in the platform.

The discontinuous reception (DRX) mechanism is commonly employed in current LTE networks to improve energy efficiency of user equipment (UE). DRX allows UEs to monitor the physical downlink control channel (PDCCH) discontinuously when there is no downlink traffic for them, thus reducing their energy consumption. However, DRX power savings are achieved at the expense of some increase in packet delay since downlink traffic transmission must be deferred until the UEs resume listening to the PDCCH. In this paper, we present a promising mechanism that reduces energy consumption of UEs using DRX while simultaneously maintaining average packet delay around a desired target. Furthermore, our proposal is able to achieve significant power savings without either increasing signaling overhead or requiring any changes to deployed wireless protocols.

Social networks are being massively utilized for targeting consumers, as entertainment channels and in personal communications, but also offer an appealing potential as a tool for learning in higher education, largely unexplored yet. This paper describes the design and implementation of a social network platform aimed at integrating informal learning processes (games, questions and answers, group ranking) in the university curricula. We present the design principles, the major software modules and the workflow of our tool, which is based on open-source software.

The IEEE 802.3az standard provides a mechanism to build energy efficient Ethernet interfaces via a low power idle mode that they can enter when there is no data to transmit. Several competing algorithms have appeared that make use of this mode to minimize energy consumption with little disruption to the traffic. Two algorithms stand out among those because of their simplicity and performance: frame transmission and burst transmission. Although these algorithms have been shown to be very efficient in simulated scenarios, there is a lack of general analytical models for their behavior. In fact, to this date, the only analyzed traffic patterns have been variants of Poisson traffic. In this paper we provide a general GI/G/1 model for energy consumption and traffic delay for both algorithms. We then develop specializations of the general model for Poisson and deterministic traffic. Finally, we validate the model with the help of both synthetic traffic and real Internet traffic traces.

In this paper, we research the feasibility of adapting the packet coalescing algorithm, used successfully in IEEE 802.3az Ethernet cards, to upstream EPON channels. Our simulation experiments show that, using this algorithm, great power savings are feasible without requiring any changes to the deployed access network infrastructure nor to protocols.

In this paper, we propose probabilistic splitting of a packet stream at the edge routers as the basic method to provide end-to-end proportional QoS to packet flows carried through an OBS network, in terms of loss probability. We argue that the only requirement that the optical transport infrastructure has to satisfy is the support for two internal burst classes with wide separation between their respective service levels. Under this condition, we show how quantifiable end-to-end per-flow guarantees can be attained without diminishing network resource usage. The scheme is analyzed theoretically and evaluated through numerical simulations, both in regular topologies (ring networks) and in a mesh network. Our results suggest that a layered approach to the problem of end-to-end QoS provisioning can be very effective when a proportional service model is offered, and that the research on sophisticated scheduling algorithms at the optical switches could be too narrow-focused for that purpose.

Network interfaces in most LAN computing devices are usually severely under-utilized, wasting energy while waiting for new packets to arrive. In this paper, we present two algorithms for opportunistically powering down unused network interfaces in order to save some of that wasted energy. We compare our proposals to the best known opportunistic method, and show that they provide much greater power savings inflicting even lower delays to Internet traffic.

The election of the best point of connection toward the network of infrastructure plays a fundamental role during the process of configuration of a nested mobile network. Thus, in this work the associated problems to the loops generation are revised during the process of connection, and a mechanism is described that permits to solve them for arrogance of routers with multiple egress interfaces, by means of a novel algorithm of loops control and the analysis of the messages of the protocol.

The Vehicular Ad-hoc NETworks (VANET) consist of a spontaneous association of a group of vehicles with wireless connection, they move and dynamically change their positions, exchanging data between each other. These networks were thinking as autonomous network segments with flat addressing schemes, however, its study has shown the benefits obtained by interconnecting them to fixed network segments and Internet. This article will revise by means of using a simulation tool, the global improvements in QoS network metrics, achieved through the use of a mutipath routing protocol. The model has a fixed network segment (Internet), two reactive gateways and two IPv6 VANET hierarchical network segements, whose nodes show a high degree of mobility, such as vehicles in an urban environment.

This paper addresses the provision of proportional differentiated services to an arbitrary number of traffic classes in terms of the class packet loss probability measured between the ingress node and the egress node of an OBS network. Our solution relies on a key idea: OBS networks consist of bufferless nodes and can therefore be regarded as a whole like a one-hop bufferless subnetwork characterized by a collection of loss probabilities computable with simple, approximate circuit switching models. Consequently, we believe it is possible to attain packet loss proportionality merely using a simple stochastic algorithm to assemble two classes of bursts, provided there exists some form of internal relative differentiation such that one of the burst classes has a much lower loss probability than the other. In order to evaluate the algorithm accuracy in attaining the proportionality, we present an analytical study of two idealized scenarios (a single link and a multi-node symmetrical network), including the effect caused by the flow dynamics of the traffic received at the ingress nodes. Our results show that, despite its simplicity, this approach is able to provide the desired proportionality over a wide range of operating conditions.

We propose a new hybrid congestion control protocol for multimedia streaming. While dominant proposals in the field have generally dealt with either pure single-rate or multiple-rate proposals, these either lack the ability to adapt to heterogeneous receivers or are too complex to be fully understood and safely deployed. Recent advances have suggested the use of hybrid approaches, but so far these approaches fail to take into account the real capabilities of the receivers. Our proposal works by sampling the allowed transmission rate of the set of receivers and adjusting the minimum transmission rate of each multicast group according to these data. Then a single-rate congestion control protocol guarantees in each group the TCP-fairness of the approach.

One of the more challenging research issues in the context of assured forwarding (AF) is the fair distribution of bandwidth among aggregates sharing the same AF class. Several studies have shown that the number of microflows in aggregates, the round trip time, the mean packet size and the TCP/UDP interaction are the key factors in the throughput obtained by the competing aggregates. Dynamic RIO (DRIO) is an interesting RIO technique suggested to improve assurances among heterogeneous flows in AF-based services. In this paper, we propose applying DRIO to aggregated traffic instead of the individual flows. Working at the aggregate level not only makes DRIO more appropriate to be used in AF networks, but also improves scalability substantially. In addition, we also mitigate some unfairness issues found in the original scheme. Several simulation experiments have been conducted to verify that the proposed improvements allow DRIO to fulfill fairness requirements more satisfactorily.

Congestion control of multicast real-time flows in the Internet has different requirements and poses further difficulties than common end-to-end unicast congestion control. In this paper we present a novel congestion control protocol, the layered datagram protocol (LDP), aimed at solving such problems. LDP has been designed to be fair against competing TCP traffic and, at the same time, exhibit smoothness in the instantaneous throughput, making it attractive for real-time multimedia transmission. Furthermore, the open-loop approach makes the protocol highly scalable, as the server load is not affected by the number of clients present in the transmission. Our simulations show that the protocol exhibits the expected TCP fairness while still being highly tunable to smooth the instantaneous throughput.