The increasing adoption of machine learning (ML)-based services has presented challenges in processing sensitive data while ensuring privacy and confidentiality. Homomorphic encryption offers a promising solution by enabling computations on encrypted data. However, applying homomorphic encryption in ML faces challenges regarding efficient structuring, arrangement, and execution of numerical operations. In this paper, we present HEJet: a framework that enables efficient and user-friendly application of neural networks with homomorphic encryption. Our framework maps sequences of numerical computations to an optimized set of instructions that are processed by compilers for homomorphic encryption. Consequently, HEJet provides user-friendly interfaces to utilize advanced neural network structures with homomorphic encryption. Evaluation results on the MNIST dataset highlight its usability and show a significant speedup in inferences between 3% and 48% compared to existing approaches. Additionally, HEJet maintains accuracy levels close to those observed on raw data.

Emerging networks, such as 5G, can indeed deliver higher capabilities compared to previous generations. Yet, there is still a need for more capacity to meet the growing demand for higher bandwidth. Meanwhile, MPTCP has emerged as a protocol that combines existing networks to increase available bandwidth through the simultaneous use of multiple connections. However, most public servers do not currently support this protocol, forcing MPTCP connections to fall back to conventional single-path TCP. As a solution, we propose deploying a MPTCP proxy server as an enabler to provide multipathing on the client side. Using this approach, we combine the bandwidth available for 5G and LTE networks to determine the protocol’s ability to improve mobile communications. In addition, we have analyzed different congestion control and scheduling mechanisms for MPTCP to find out which are suitable for various situations and may yield significant beneficial results.

For secure remote processing, homomorphic encryption can be used. It allows operations to be performed on encrypted data. If the processor needs the result’s plaintext, it relies on the data owner to perform the decryption. This poses a vulnerability since a malicious data owner could inject a self–selected value. Moreover, it sees the plaintext result, breaching its confidentiality. In this paper, we propose a solving approach for the CKKS homomorphic encryption scheme. Using a commutative property, the data owner removes its encryption, while a processor’s encryption reinforces integrity and confidentiality. In an evaluation, we demonstrate feasibility regarding computational effort, transmitted data size, and introduced error.

QUIC aims to become a general-purpose transport protocol, and numerous implementations of the QUIC protocol already exist. Earlier evaluations often examined QUIC in conjunction with HTTP/3.0 or focused on latency metrics. The measurement studies in this paper focus on actual QUIC implementations with respect to their ability to achieve high sustained throughput in network scenarios with data rates of 10 Gbit/s. We compare six popular QUIC implementations developed in different programming languages with TCP. Our findings show significant performance improvements in several QUIC implementations compared to prior evaluations. However, it is not a homogeneous picture, as current QUIC implementations often behave quite differently. We observed that in environments with low RTTs or an increased number of packet losses, most of the surveyed QUIC implementations struggle unexpectedly and cannot compete with TCP regarding sustained throughput performance.

The in-depth consideration of security aspects in modern web infrastructures has become essential to stay competitive. In this context, continuous authentication is a promising approach to prevent the misuse of digital identities. To this end, machine learning (ML) models are well suited to analyze user behavior and to detect anomalies, due to their ability to identify complex patterns and trends that usually cannot be reflected by static rule-based approaches. However, the training of powerful ML models requires large amounts of data, which are often not available within a single organization. Consequently, a federated training of these models by cooperating organizations offers a promising solution, but leads to concerns about coordination, regulations, and quality assurance. To tackle these challenges, we present an approach that combines three research areas: (1) the establishment of continuous user authentication based on (2) a ML model trained by an organized peer-to-peer federation involving different organizations that is underpinned by (3) federated data governance ensuring regulatory compliance and quality of resulting artefacts.

As password-based authentication fails to provide adequate security for online activities such as financial transactions, additional authentication factors are required. Such factors should provide both ease of use and an adequate level of privacy protection, while being easy to implement, operate and maintain, even for applications with thousands of users. As an approach meeting these requirements, we outline Scalable and Privacy- Preserving Continuous Authentication (SPCAuth). SPCAuth determines risk levels for actions that should be authenticated, without requiring explicit user interactions. It analyzes different aspects of user behavior by means of machine learning methods, while preserving the privacy of the affected individuals. SPCAuth trains only a single model per aspect of user behavior being considered, based on observations of all users, ensuring scalability and increasing accuracy for users with infrequent activities. In two experiments, we have confirmed that this key concept enables a scalable and accurate user authentication.

Most approaches for providing incentives for resource sharing on energy-constrained mobile devices rely on some kind of payments, secured by means of public key cryptography. Checking the validity of such payments creates substantial energy overhead itself. We evaluate the energy overhead of an incentive system based on hash coins. Therefor we created a prototype of a hash coin based incentive system in a wireless ad hoc network and conducted energy measurements. The results are compared to solutions based on public key cryptography. Using the knowledge about the energy consumption of these approaches helps us to chose the best suited incentive system and reasonable system parameters for an incentive system in a wireless ad hoc network.

This paper contributes a protocol that measures underlay connectivity properties of nodes in an overlay network. The protocol autonomously assigns a unique random identifier to each subset of nodes that can communicate transitively using arbitrary underlay routing protocols (e.g., IPv4, IPv6, RFCOMM), denoted connectivity domain. We denote nodes in more than one connectivity domains relays, as they can forward messages between connectivity domains. This information is useful for overlay construction, as possible overlay neighbors must be in the same connectivity domain while relays enable communication between nodes in different connectivity domains. The general problem of identifying maximal connectivity domains is equivalent to the problem of finding all maximal cliques. We show that it can be efficiently approximated in an overlay setting with only partial knowledge of the network connectivity graph. The key observation is that in such scenario once all we discovered all relays remaining nodes can agree on a unique identifier per connectivity domain. We show by theoretical analysis and extensive simulations that the approach is accurate, scalable, resilient, and reactive to connectivity changes in the underlay network.

Because introducing new protocols and services, e. g., multicast or IPv6, on a global scale has proven to be difficult for various reasons, Peer-to-Peer (P2P) systems have become a method of choice for introducing innovation in the Internet. Such systems create Overlay Networks consisting of logical connections between end-systems at the edge of the Internet, do not require any supporting infrastructure, are scalable, and can be deployed using self-organization. However, the evolution of the Internet dramatically increases heterogeneity: NAT boxes establish different address spaces in different parts of the network. The partial transition from IPv4 to IPv6 leads to different protocols in the network layer and it is unsure whether all networked devices will use IP at all. These trends hinder direct connectivity between endsystems and, thus, make overlay deployment challenging. In this paper we survey on overlay approaches that can cope with these obstacles by providing themselves the functionality for homogeneous and persistent addressing and routing across heterogeneous networks. We discuss general approaches and trade-offs by example systems. Motivated by the lessons learned from these systems we present the Abstraction Base ariba, that implements shortest path routing across heterogeneous networks, maintains few neighborhood information, and requires low overhead for updates induced by network dynamics and node mobility.

Presence technology enables users of an instant messaging (IM) system to determine if their contacts are online and ready to communicate. In this paper, we propose an effective approach for the proactive dissemination of frequently changing presence information in highly partitioned mobile, wireless networks with IEEE 802.11 technology. Although communication techniques for intermittently connected networks have been extensively studied in the field of delay tolerant networking, the fact that presence information is highly delay sensitive requires a thorough revision of these techniques. To this end, we use discrete-event simulation based on a high-level stochastic model of the IM system to compare different approaches for disseminating presence information in terms of sustained consistency (i.e., fraction of time presence information is in a coherent state) and traffic requirements. Building upon the outcome of the simulation study, we propose the system for presence information exchange by epidemic dissemination (SPEED). Results of a detailed ns-2 simulation study show that SPEED outperforms an approach based on optimized flooding by up to 20% in terms of sustained consistency for low node density and saves up to 48% of control traffic for medium to high node density

Epidemic algorithms have recently been proposed as an effective solution for disseminating information in large-scale peer-to-peer (P2P) systems and in mobile ad hoc networks (MANET). In this paper, we present a modeling approach for steady-state analysis of epidemic dissemination of information in MANET. As major contribution, the introduced approach explicitly represents the spread of multiple data items, finite buffer capacity at mobile devices and a least recently used buffer replacement scheme. Using the introduced modeling approach, we analyze seven degrees of separation (7DS) as one well-known approach for implementing P2P data sharing in a MANET using epidemic dissemination of information. A validation of results derived from the analytical model against simulation shows excellent agreement. Quantitative performance curves derived from the analytical model yield several insights for optimizing the system design of 7DS.

This paper characterizes the query behavior of peers in a peer-to-peer (P2P) file sharing system. In contrast to previous work, which provides various aggregate workload statistics, we characterize peer behavior in a form that can be used for constructing representative synthetic workloads for evaluating new P2P system designs. In particular, the analysis exposes heterogeneous behavior that occurs on different days, in different geographical regions (i. e., Asia, Europe, and North America) or during different periods of the day. The workload measures include the fraction of connected sessions that are passive (i. e., issue no queries), the duration of such sessions, and for each active session, the number of queries issued, time until first query, query interarrival time, time after last query, and distribution of query popularity. Moreover, the key correlations in these workload measures are captured in the form of conditional distributions, such that the correlations can be accurately reproduced in a synthetic workload. The characterization is based on trace data gathered in the Gnutella P2P system over a period of 40 days. To characterize system-independent user behavior, we eliminate queries that are specific to the Gnutella system software, such as re-queries that are automatically issued by some client implementations to improve system responsiveness.

In this paper, we present a comprehensive performance study of least recently used and least frequently used with dynamic aging as traditional replacement schemes as well as for the newly proposed schemes greedy dual size and greedy dual. The goal of our study constitutes the understanding how these replacement schemes deal with different web document types. Using trace-driven simulation, we present curves plotting the hit rate and byte hit rate broken down for image, HTML, multi media, and application documents. The presented results show for the first workload that under the packet cost model Greedy Dual outperforms the other schemes both in terms of hit rate and byte hit rate for image, HTML, and multi media documents. However, the advantages of Greedy Dual diminish when the workload contains more distinct multi media documents and a larger number of requests to multi media documents.

In this paper, we present the concept of passive distributed indexing, a general-purpose distributed search service for mobile file sharing applications, which is based on peer-to-peer technology. The service enables resource-effective searching for files distributed across mobile devices based on simple queries. Building blocks of PDI constitute local broadcast transmission of query- and response messages, together with caching of query results at every device participating in PDI. Based on these building blocks, the need for flooding the entire network with query messages can be eliminated for most application. In extensive simulation studies, we demonstrate the performance of PDI. Because the requirements of a typical mobile file sharing application are not known-or even do not exist at all-we study the performance of PDI for different system environments and application requirements. We show that due to the flexible design PDI can be employed for several kinds of applications.