Smart cities, leveraging IoT technologies, are revolutionizing the quality of life for citizens. However, the massive data generated in these cities also poses significant privacy risks, particularly in de-anonymization and re-identification. This survey focuses on the privacy concerns and commonly used techniques for data protection in smart cities, specifically addressing geolocation data and video surveillance. We categorize the attacks into linking, predictive and inference, and side-channel attacks. Furthermore, we examine the most widely employed de-identification and anonymization techniques, highlighting privacy-preserving techniques and anonymization tools; while these methods can reduce the privacy risks, they are not enough to address all the challenges. In addition, we argue that de-identification must involve properties such as unlikability, selective disclosure and self-sovereignty. This paper concludes by outlining future research challenges in achieving complete de-identification in smart cities.

Cloud-network slicing is a promising approach to serve vertical industries delivering their services over multiple administrative and technological domains. However, there are numerous open challenges to provide end-to-end slices due to complex business and engineering requirements from service and resource providers. This article presents a reference architecture for the cloud-network slicing concept and the practical realization of the slice-as-a-service paradigm, which are key results from the Novel Enablers in Cloud Slicing (NECOS) project. The NECOS platform has been designed to consider modularity, separation of concerns, and multi-domain dynamic operation as prime attributes. The architecture comprises a set of interworking components to automatically create, manage, and decommission end-to-end cloud-network slice instances in a lightweight manner. NECOS orchestrates slices at runtime, spanning across core/edge data centers and wired/wireless network infrastructures. The novelties of the multi-domain NECOS platform are validated through three proof-of-concept experiments: (i) a touristic content delivery service slice deployment featuring on-demand virtual infrastructure management across three countries on different continents to meet particular slice requirements; (ii) intelligent slice elasticity driven by machine learning techniques; and (iii) market-place-based resource discovery capabilities.

Aims/hypothesis: There are five systems to stratify the risk for the development of a diabetic foot ulcer (DFU). This study aimed to prospectively validate all of them in the same cohort of participants to allow their direct comparison.Methods: A retrospective cohort study was conducted on all patients with diabetes but without an active DFU attending our podiatry section (n=364) from January 2008 to December 2010. Participants' characteristics and all variables composing the stratification systems were assessed at baseline. Follow-up was performed for 1 year or until DFU occurred.Results: Participants had a mean age of 64 years; 99.7% had type 2 diabetes and 48.6% were male. Median follow-up was 12 months (1–12) during which 33 subjects (9.1%) developed a DFU. Age, diabetes duration, foot deformity, peripheral vascular disease, diabetic peripheral neuropathy, previous DFU, and previous lower extremity amputation were associated with DFU occurrence. All systems presented greater DFU occurrence frequency as the risk group was higher (χ²,P<0.001) and showed good diagnostic accuracy values, especially negative predictive value (≥95%) and area under the receiver operating curve (≥0.73). The lowest performance concerned positive predictive value (≤29.5%).Conclusions/interpretation: All the currently available stratification systems show high accuracy to detect which patients will develop a DFU with no significant differences among them. Therefore, for diabetic foot screening and resource allocation, it would be desirable to have a single unified system, combining the available systems, prospectively validated in a multicenter context and testing the inclusion of novel predictive variables' pertinence.

The design of efficient elasticity control mechanisms for dynamic resource allocation is crucial to increase the efficiency of future cloud‐network slice‐defined systems. Current elasticity control mechanisms proposed for cloud‐ or network‐slicing, only consider cloud‐ or network‐type resources respectively. In this paper, we introduce the elaSticity in cLOud‐neTwork Slices (SLOTS) which aims to extend the horizontal elasticity control to multi‐providers scenarios in an end‐to‐end fashion, as well as to provide a novel vertical elasticity mechanism to deal with critical insufficiency of resources by harvesting underused resources on other slices. Finally, we present a preliminary assessment of the SLOTS prototype in a real testbed, revealing outcomes that suggest the viability of the proposal.