Deep image compression algorithms based on Generative Adversarial Networks (GANs) are a promising direction to address the strict communication bandwidth limitations commonly encountered in IoT sensor networks (e.g. Low Power Wide Area Networks). However, current methods do not consider that the sensor nodes, which perform the image encoding, usually only offer very limited computation and memory capabilities, e.g. a resource-constrained tiny device such as a micro-controller. In this paper, we propose the first tiny generative image compression method specifically designed for image compression on micro-controllers. We base our encoder on the well-known MobileNetV2 network architecture, while keeping the decoder side fixed. To cope with the resulting asymmetric design of the compression pipeline, we investigate the impact of different training strategies (end-to-end, knowledge distillation) and integer quantization techniques (post-training, quantization-aware training) on the GAN-training stability. On the Cityscapes dataset, we achieve a compression performance that is very close to the state-of-the-art, while requiring 99% less SRAM size, 97% smaller flash storage and 87% less multiply-add operations. Our findings suggest that tiny generative image compression is particularly well suited for application-specific domains.

Online Social Networks (OSNs) are the core of most communications nowadays, leading to possibly sensitive information exchange. Privacy is an important building block of free societies, and thus, for OSNs. OSNs function as group communication systems and can be build in centralised and distributed styles. Privacy can be achieved in distributed systems as all participants contribute to privacy. Peer-to-peer-based group communication systems achieve this privacy improvement partially, at the cost of additional messaging overhead. In this paper, we introduce ant colony optimisation to reduce the messaging overhead of anonymous communication systems, bridging the gap between privacy and efficiency. We apply our adapted privacy sensitive ant colony optimization to improve routing paths by encouraging re-usage and aggregation. Our first results indicate a 9-13% lower messaging overhead compared to the state of the art, while maintaining privacy.

Defending key network infrastructure, such as Internet backbone links or the communication channels of critical infrastructure, is paramount, yet challenging. The inherently complex nature and quantity of network data impedes detecting attacks in real world settings. In this paper, we utilize features of network flows, characterized by their entropy, together with an extended version of the original Replicator Neural Network (RNN) and deep learning techniques to learn models of normality. This combination allows us to apply anomaly-based intrusion detection on arbitrarily large amounts of data and, consequently, large networks. Our approach is unsupervised and requires no labeled data. It also accurately detects network-wide anomalies without presuming that the training data is completely free of attacks. The evaluation of our intrusion detection method, on top of real network data, indicates that it can accurately detect resource exhaustion attacks and network profiling techniques of varying intensities. The developed method is efficient because a normality model can be learned by training an RNN within a few seconds only.

Visualisations are often used to communicate trust-worthiness to end users. Showing a number of stars, for example, is a well-known practice in e-commerce applications to communicate the quality of a product or service. Many products or services also have their quality - in terms of trustworthiness - described along more than one dimension, so that not only an overall trust score has to be communicated, but multiple scores, one for each dimension. Current visualisations of such a multicriterial trustworthiness are often based on the display of multiple individual star-like interfaces - a practice that offers room for improvement with regard to intuitive understanding of the displayed trust information. In this paper, we present T-Viz, a trust visualisation based on radar plots and pie charts. T-Viz concurrently shows multiple trust scores, one for each dimension, along with an aggregated trust score. Moreover, T-Viz also shows a reliability measure for every trust score graphically, in the form of a certainty score. The evaluation results from a pilot study with eleven participants indicate that T-Viz is an intuitive, comprehensible and clear interface. It succeeds at visualising and communicating multicriterial trust scores under uncertainty in one, easy to understand, graphical representation.

Common conceptualizations of trust concern expectations about the future behavior of an entity. We focus on expectations about fair future behavior: this aspect was insufficiently formalized to date, prohibiting formal reasoning about fairness. In this paper, we extend our previous binary fairness model to a gradual one. We demonstrate the capabilities of gradual fairness in two scenarios: "distributing revenue in open-ended funds" and "e-business retailing". In both scenarios, independent partners decide on business collaboration under incomplete information. Furthermore, we contribute an approach to determining an optimal distribution under different types and grades of fairness expectations.

Trust and reputation systems (TRS) require a composition of individual methods for prediction, selection and feedback/data generation in order to fully realise their potential. Building on recent work identifying common aspects of TRS, this paper presents a prototype of a software deployment tool for assembling TRS from individual modules. This tool is considered as a precursory support step to the systematic testing and comparison of the various facets that constitute the trust assessment process. This will support future research and provide an engineering perspective to the development of new TRS, by providing easy deployment of TRS for testing.

The fusion of trust relevant information provided by multiple sources is one of the major challenges of trust establishment, which in turn is a key research topic in the growing field of cloud computing. We present a novel fusion operator for combining information from different sources, representing propositions under uncertainty. The operator especially extend the state-of-the-art by explicitly considering weights and the handling of conflicting dependent opinions. We provide a use case that demonstrates the applicability of our approach and shows the capability of the novel operator to a more reliable and transparent assessment of the trustworthiness of cloud providers.

Interactions between individuals are inherently dependent upon trust, no matter if they occur in the real world or in cyber communities. Over the past years, proposals have been made to model trust relations computationally, either to assist users or for modeling purposes in multi-agent systems. These models rely implicitly on the social networks established by participating entities (be they autonomous agents or internet users). However, state-of-the-art trust frameworks often neglect the structure of those complex networks. In this paper, we present a new approach allowing agent-based trust frameworks to leverage information from so-called weak ties that would otherwise be neglected. An effective and robust voting scheme based on an agreement metric is presented and its benefit is shown through simulations.