Causal Neural Network models have shown high levels of robustness to adversarial attacks as well as an increased capacity for generalisation tasks such as few-shot learning and rare-context classification compared to traditional Neural Networks. This robustness is argued to stem from the disentanglement of causal and confounder input signals. However, no quantitative study has yet measured the level of disentanglement achieved by these types of causal models or assessed how this relates to their adversarial robustness.

We aim to provide a solution for the secure identification of sensitive medical information. We consider a repository of de-identified medical data that is stored in the custody of a Healthcare Institution. The identifying information which is stored separately can be associated with the medical information only by a subset of users referred to as custodians. This paper intends to secure the process of associating identifying information with sensitive medical information. We also enforce the responsibility of the custodians by maintaining an immutable ledger documenting the events of such information identification. The paper proposes a scheme for constructing ledger entries that allow the custodians and patients to browse through the entries which they are associated with. However, in order to respect their privacy, such traversal requires appropriate credentials to ensure that a user cannot gain any information regarding the other users involved in the system unless they are both involved in the same operation.

In a sensor-cloud infrastructure scheduling and allocation of resources are challenging tasks. Unlike computational cloud infrastructure, sensor may have varying capabilities and contexts of services. Moreover, some sensors may be available for a fixed time duration periodically. Depending on some dynamic factors, performance of a sensor may also vary. Such non-uniform performance may be predicted from previous observations. This article proposes novel scheduling algorithms for heterogeneous resources having varying capabilities, contexts of usage and non-uniform performance over time. Simulation results show that the algorithms perform efficiently.

In this paper we focus on solving the problem of scheduling heterogenous resources in Sensor Cloud Infrastructure. We consider that the sensing resources are periodically available and they vary, not only by their types, but also by their capabilities. It is also assumed that the performance of the resources are time varying and non-uniform. The requests specify the type and the capabilities it requires, and a time span it is interested in. There may not be a single resource that is available through the requested time span. Our objective is to schedule one or more such periodically available resources, that match the requested type and capabilities while minimizing resource migrations and maximizing the overall performance of the service.

In this paper, we propose a sensor-cloud environment which incorporates remotely located physical sensor devices in the generic resource family along with other resources like CPU, memory etc. The environment enables ubiquitous and on-demand access to virtual sensors which are abstractions of physical devices with enhanced capabilities. An architecture of the sensor-cloud environment is discussed in this paper that spans multiple infrastructures hosting local and remote resources. Responsibilities of each of the components in the sensor-cloud architecture are defined and algorithms for provisioning these components are discussed along with some proposed enhancements with respect to our earlier work. As forwarding data traffic is an important task, a high performance I/O redirection algorithm is also proposed to deliver sensed traffic to multiple instances of virtual sensors that is accumulated from one or multiple physical sensors. Performance of the I/O redirection algorithm is observed and the experimental results are presented in this paper.