The ubiquitous presence of Internet-of-Things devices represents a menace to cybersecurity since their low computational power does not allow classic countermeasures techniques. In recent years, anomaly detection using Machine Learning has acquired popularity among researchers; however, current datasets in literature are sub-optimal, making exhaustive benchmarks challenging to create. Moreover, the current proposed ML solutions do not consider the necessity of adapting to IoT networks that mutate fast. We propose a Continuous Learning approach to anomaly detection on IoT networks: frequent model retraining and the exploitation of previous knowledge allow a classifier to quickly adapt to new network configurations. We deployed this approach firstly by defining a benchmark methodology for ML algorithms that helps train and test models for anomaly detection in IoT networks. This phase allowed us to identify the best-performing algorithm for the task. Then, we employed Semi-supervised Learning techniques to deploy the continuous part concretely. Our results show that applying Continuous Learning can progressively improve anomaly detection performance in key metrics such as Recall (up to 75%) and Roc AUC (up to 33%).

Abstract Wearable technology is no longer science fiction. Thanks to the growing capability in the production chain to miniaturize complex electronics, a wide variety of gadgets that can be worn or included in dresses and accessories have emerged. These smart gadgets can collect data about the physical condition of the user and/or the environment providing the basis for innovative and valuable services. The main goal of this paper is to assess this context through field experiments undertaken in a testbed comprised of sensing hardware deployed on open source boards such as Arduino. Moreover, coupled with the sensing tier, we propose a proof‐of‐concept deployment architecture enabling a wide range of wearable sensors to collect and transmit data to a logically centralized unit.

Even if the World Wide Web is one of the main content and service providers, unfortunately, these contents and services are not really available for everyone. People affected by impairments often have difficulties in navigating Web pages for a wide range of reasons. In this paper, we focus on people affected by dyslexia. These users experience difficulties in reading acquisition, despite normal intelligence and adequate access to conventional instruction. For this reason, we have created Help me read!, a Chrome extension that allows to change many features of a Web page. Furthermore, it allows to isolate and enlarge one word at a time. This feature is crucial as it allows people with dyslexia to focus on each single word, thus overcoming one of their main difficulties.

Is the World Wide Web for everyone? Long story short: no and unfortunately it is not only a matter of infrastructures, level of instruction or economic conditions. People affected by visual impairments have often difficulties in navigating web pages for a wide range of reasons. One of the biggest obstacles nowadays is the use of CAPTCHAs, powerful tools against bot attacks but also potential virtual barriers for the aforementioned category of users. In this paper we tested various categories of CAPTCHAs with people affected by visual impairment and not, to understand how discriminatory can be these cybersecurity measures.

The fruition of multimedia contents has been revolutionized by the spread of broadband technology. Furthermore, the advent of the cloud is now moving the computational load, once prerogative of domestic devices, to remote servers. This is bringing achievements that were unthinkable only few years ago, like the provisioning of new multimedia services through the Internet. To guarantee a proper user experience, a low E2E latency and a high bandwidth are both essential. Unfortunately, this is difficult to accomplish in a shared and best-effort network like the Internet. We need to focus on the study and analysis of modern network protocols to identify viable solutions for a suitable user experience. This is especially the case of QUIC, a recent transport protocol able to reduce significantly some issues mentioned above. Our aim is to explore QUIC and its integration into a Smart Access Point, enabling a better management of heterogeneous co-existing media flows.

Cloud-based interactive multimedia applications such as virtual games and video streaming are gaining high popularity. However, giving the high bandwidth consumption, the remote execution can negatively impact the quality of the multimedia traffic. In such a realm, data travel different communication networks from the cloud to the final users crossing the last meters the home's access point (AP). In such a scenario, the quality-of-service (QoS) support is a challenging task, particularly in the home network environment, with heterogeneous applications simultaneously running and consuming the available bandwidth. To address this issue, we propose ReiLeCS, a Reinforcement Learning-based Controller and Scheduler for interactive multimedia traffic in Home Area Networks (HAN). Through reinforcement learning and the maximization of a reward function, it enables the AP to schedule the arriving multimedia traffic from the cloud according to their required QoS. Simulation results using real multimedia traffic conditions demonstrate that ReiLeCS achieves better performances compared with existing packet scheduling policies.

Although much has been done in the field of accessibility, the World Wide Web is not yet for everyone. Many Internet users often find difficulties in navigating web pages. This is particularly true for people with visual impairments and blind people. In this paper, we present a review of the state of the art of the application of accessibility rules to web pages and describe which are the common problems that people with sight impairments must endure when surfing the Internet; to this aim, we use information collected during an interview with a blind user who is an expert in navigating the web using a screen reader.

Wearable devices are constantly expanding in number and functionalities within both the IT market and the fashion market. Smart watches have become very popular and have moved from a tool only showing information related to the smartphone to a device able to monitor and display a vast plethora of data, including data related to the owner's health (e.g., hearth rate, body temperature, steps per day, etc.). This huge amount of data that wearables in the world can generate could be further elaborated to become even more useful. Yet, designing effective solutions to collect, store and elaborate sensors data is a challenging task. To this aim, in this paper, we present a framework to model a wearable sensors network in a simple and flexible in order to adapt to different scenarios.

One of the most profitable and emerging high-tech markets is the one that involves games for PC and consoles; a market that is growing year by year and that is rapidly evolving. Several big game corporations are implementing now the remote video streaming through the domestic network between two devices; a feature that regards not only PCs but also consoles. This paper investigates one of the main services of this type, the Steam In-Home Streaming, to analyze the generated network traffic when considering different types of game. Gathered data can help to understand whether this type of service can actually be supported by today's regular Internet connections.

The market penetration of in-home entertainment systems is playing a crucial role in the diffusion of interactive multimedia services. In this context, thin-client game systems have received a lot of attention, shifting the computational burden of the game play toward the server while the client renders a fat video stream. This paper proposes an analysis of the network characteristics of the Remote Play for PlayStation 4 and PlayStation Vita, providing some insights on the network dynamics of this system under different game types. The measurements involve video games of different genres, analyzing the bitrates, the packet size, capturing the dynamics of the whole streaming session. Results indicate that the service generates a large quantity of big packets downstream and smaller and lesser packets upstream. Gathered data can help to understand if this type of service could be offered through the today's normal Internet connections.

This paper investigates one of the main game streaming services, the Steam In-Home Streaming, to analyze the generated network traffic. Gathered data can be used to model this type of traffic in simulations and to determine whether it could be supported by Internet connections around the world.