From scattered data to actionable knowledge: flexible cyber security reporting in the military domain

Numerous cyber situational awareness models have been proposed in recent years. Yet, one of the main challenges still remains mostly unsolved, which is what information sources contribute to the process for establishing cyber situational awareness and how is relevant information collected. While previous scientific works focused on situational awareness models and decision support based on common operating pictures, ingesting and maintaining a consistent data basis for the cyber domain has rarely been studied in detail. However, this is crucial when data distributed across different systems need to be collected, vetted, correlated, de-duplicated, enriched and finally stored as a basis for flexible cyber security reporting. In this paper, we design an approach and a data model that enable to ingest and store the essential information from disparate organizational units and act as a basis for the flexible creation of cyber security reports. We describe the application of this approach and model in a case study together with the Austrian Ministry of Defense (MoD), in which we surveyed existing data sources and transfer paths and rated the applicability of the CCOP data model and accompanying processes in course of a proof-of-concept implementation.


Introduction
Situational awareness is a cornerstone of justified decision making. It deals with the perception of the element in the environment within a volume of time and space, the comprehension of their meaning, and the projection of their status in the near future [1]. Cyber situational awareness [2] specifically deals with building awareness in a cyber environment, where in contrast to physical environments, time and space have different meanings. Ministry of Defence Austria, Vienna, Austria include common elements, i.e., relevant to a wide range of actors that help to establish situational awareness and to act upon accordingly. This means that information must be collected, disseminated, and represented in a way that is useful to several stakeholders in order to get a common understanding of an ongoing situation [3]. A definition of a COP could be any actively selected information that is useful to multiple stakeholders with a common overarching mission [3]. A cyber common operating picture (CCOP) aims to achieve to create this common view for the cyber domain and is an important pillar of establishing cyber situational awareness. Therefore, a CCOP provides situational awareness despite cyberspace's largely opaque nature, enhances a leader's ability to make quicker critical decisions [4]. From this definition, it can be inferred that the audience of CCOPs in the military domain is the command level, with the CCOP focusing on the current state of the military organization's cyber assets. The dissemination of a CCOP can be achieved in different ways, tool-based or in paper form. This paper focuses on a tool-based approach using reports.
Dozens of situational awareness models have been discussed in recent decades. For the cyber domain, several of these have been adopted. Yet, one of the main challenges still remains mostly unsolved, which is what information sources contribute to the process for establishing cyber situational awareness and how is relevant information collected. While hundreds of scientific works focus on situational awareness models and decision support based on common operating pictures [5], ingesting and maintaining a consistent data basis for the cyber domain has rarely been studied in detail. However, this is crucial when data distributed across different systems need to be collected, vetted, correlated, de-duplicated, enriched and finally stored as a basis for flexible cyber security reporting.
A cyber situation center (CSC) is an organizational unit that centrally perceives and interprets the security status of ICT-based technical infrastructures of one or more organizations, and creates CCOPs as the basis for flexible cyber security reporting. In addition to commercial use in large enterprises, cyber situation reports (CSR) are created in national cyber situation centers for a wide variety of stakeholders.
Unfortunately, there is hardly any publicly available, international literature on the creation of cyber situation reports from highly dispersed data at the state level, specifically for the military domain. We therefore investigate this aspect in this paper. For that purpose, we carefully distinguish the notions of information source, information class and transmission path, which are often confused. For instance, e-mail is not a source, it is rather a technical transmission path, since the actual information comes however from a technical audit or a commercial threat intelligence feed. Different types, i.e., classes, of information exist, including information about incidents, events, assets, new risks, audit findings, threat intent or threat actors. We aim to survey the different data sources and investigate how they are utilized to enable cyber security reporting in context of a military case study. Notice, the detailed implementation and choice of technologies is not in scope of this paper, which merely focuses on the methodological aspects only.
The contributions of this paper are: -Methodology to cyber common operating pictures (CCOP) : We introduce the key properties of CCOPs and outline a methodology together with a process to establish them. -CCOP data model: We specifically focus on a concept to ingest and store the essential information from disparate organizational units into a common model, which acts as a basis for the flexible creation of cyber security reports. -Survey and case study with the Austrian MoD: We describe the results of a case study together with the Austrian Ministry of Defense (MoD), in which we surveyed existing data sources and transfer paths and rated the applicability of the CCOP data model and accompanying processes in course of a proof-of-concept implementation.
The remainder of the paper is organized as follows. Section 2 contains background and related work. Section 3 outlines a methodology for the creation of CCOPs, while Sect. 4 shows the underlying data model, containing the various entities and their relations to capture all the relevant information for the creation of CCOPs. Section 5 reports the findings of a detailed case study carried out 2021 with the Austrian MoD that investigated the application of the introduced CCOP model. Finally, Sect. 6 concludes the paper.

Background and related work
In order to understand the rationale behind this paper, it is important to identify what the research objectives are. The main driver was the streamlining and optimization of processes to generate cyber situational awareness for MoDs, with a special focus on the Austrian MoD. Therefore, it is necessary to create the foundations of automation capabilities, which encompasses the definition of a homogenized data model tailored to the needs of the MoD. This is the basis for developing a tool that will assist the experts within the Austrian MoD in creating situational awareness by the means of a CCOP for the command level.

Terms and definitions
Many definitions of situational awareness already exist and the use of the term can be traced back to the First World War. The early definitions-for example the definition of Hamilton [6], Harwood [7] or Billings [8]-focus on human aspects in different crisis situations. They describe situation awareness as cognitive perception and knowledge transfer. One of the most frequently used definitions of situational awareness is from Endsley [9]. In the literature, decision makers try to understand and evaluate the situation (i.e., to build up situational awareness) in order to then decide which measures should be taken. Munir et al. [10] have defined and explained situational awareness, which is an integral part of command and control, from a military and air force perspective. The authors break down situational awareness into three distinct parts, namely perception (perception of the status, attributes and dynamics of the entities), comprehension (understanding of entities) and projection (estimation of the status of entities in the future).
Furthermore, in the military domain, the information environment is considered a part of the operational environment, thus influencing decisions of the command level and the employment of capabilities [11]. Gaining a complete picture of the cyber situation is therefore crucial. The term cyber sit-uational awareness (CSA) transports these basic ideas into the cyber domain. Models for (cyber) situational awareness already exist, including the Situational Awareness Model by Endsley [9], Situational Awareness Model by Okolica et al. [12], and Situational Awareness Reference Model by Tadda and Salerno [13]. The authors of [14] propose a methodology to set up CSA measurement experiments. However, this is done within the context of simulated cyber defense exercises. Focusing on CSA in the networking domain, the authors of [15] propose CRUSOE, an extensible layered data model, which consists of layers describing a system. For example, a layer may describe mission statements, detection and response capabilities or a network topology. In [16], the authors aim to develop a taxonomy of CSA tools and components for network security, while following a three-level model (perception of data and elements of an environment, comprehension of a situation and projection of future states and events). In summary, the literature shows that the models attempt to depict the human cognitive processes in the creation of situational awareness [2]. However, these processes do not yet have a high degree of maturity and the models mentioned in the literature only depict partial aspects.
A cyber common operating picture (CCOP) [4], derived from the concept of general COPs [5], captures the actual cyber security status of an organization's systems for a wide variety of stakeholders, including security staff and decision makers. A CCOP is expected to support security staff and managers to assess a situation and properly respond to risks and security threats. In order to do that many different types and classes of data and information must be collected, aggregated and interpreted. The cyber attack information system (CAIS) framework by Skopik et al. [17] suggests that the national cyber situation center develops nation-wide situational awareness from the information collected, and recommendations for action for effective response to security incidents, as well as long-term recommendations for increasing resilience. Other works focus more on the attribution process and the question to what extent cyber attacks can be associated with known actor groups [18].

Operational, tactical and strategic levels
CCOPs can have different focus and have diverse functions for the addressees. In order to understand the requirements formulated in Sect. 3.1, it is of great importance to understand the three target levels of a CCOP, namely operational, tactical and strategic. One major concept of modern military theory is the division of war into strategic, operational, and tactical levels. Here, the strategic level focuses on defining and supporting national policy, the operational level focuses on the use of military forces in a theater of operations and the tactical level is concerned with the details of how battles are conducted and is sensitive to the changing parameters of an ongoing engagement [19]. Similar concepts can also be found in other domains for instance in decision making [20]. A good definition, tailored to the military domain is given by Ardil [21]: Strategic military decisions affect all or most of the organization and directly contribute to the achievement of the common goals of the organization. Tactical military decisions serve the implementation of strategic decisions. Operational military decisions are focused on day-to-day operations and have a have a short-term horizon. Also relevant for this paper, the three terms are used in information security management. White [22] postulates three questions that must be addressed by the different levels of security management: Strategic management answers the question "why do security enterprise problems exist?" and is concerned with developing security policies. Tactical management answers the question "how are security problems mitigated?" and deals with the development of security systems which enforce the security policies. Operational management answers the question "what security procedures and practices are to be utilized?" and focuses on the use concrete security tools and technologies.

Standards and building blocks for a military-oriented CCOP
On a technical level, standards, taxonomies and ontologies already exist to represent relevant information for generating a CCOP in a structured way. The structured threat information expression (STIX) [23] specification from OASIS defines some of these information classes, including incident, event, asset, campaign, malware, threat actor and so on. Others, such as the NISTIR 8138 vulnerability description ontology [24] is more vulnerability-focused, than threat-centric. Also focusing on vulnerabilities, the Common Vulnerability Scoring System (CVSS) is a commonly used framework communicating the characteristics and impacts of vulnerabilities [25]. This rating approach has also been endorsed by the US Department of Defense [26]. The NIST Special Publication 1800-5: IT Asset Management [27] on the other side, contains valuable hints on IT asset management, and the ENISA Reference Incident Classification Taxonomy [28], elaborates, among others, on methods to capture and categorize incident information. Guidelines for assessing information in terms of relevance and reliability of sources are provided by the US Army Field Manual 2-22.3 [29]. Although none of these standards can be used as the exclusive basis for a CCOP in the military environment due to their limited scope on partial aspects, elements can be used or integrated as building blocks into a harmonized data model. The (technical) information sources can be anything from simple RSS web feeds, websites or (internal) reporting and ticketing systems to highly specialized information systems such as configuration management databases (CMDBs) or dedicated threat sharing platforms. For instance, MISP 1 is a threat intelligence platform, which allows for sharing indicators of compromise (IOC) within and between organizations. MISP also defines taxonomies, for tagging, classifying and organizing information. Collaborative cyber threat intelligence [30] deals with solutions to enable sharing of security-relevant information, including incident and vulnerability information, across organizational boundaries. While this work is relevant to some extent, we rather focus on sharing of such information within the same organization to make it centrally available for flexible cyber security reporting to decision makers. Help desk software, such as Zendesk, Spiceworks or Jira Service Desk, can also serve as an important information source for the status of different services within the own organization.

Tools for supporting the creation of a CCOP
Once all the relevant information for CCOPs have been identified, preferably automated mechanisms help to collect this information from diverse sources through technical sources. Ontology-based threat intelligence collection frameworks have been proposed, e.g., [31]; the solution intelMQ 2 is one example of such an automated crawler and collection framework that helps to collect, categorize, de-duplicate, transform and process incoming information in a structures manner. Another project focusing on receiving, enriching and redistributing information is AbuseHelper 3 . Although there have been no updates in the last three years (as of May 2022), AbuseHelper is very well-known for the automated collection of abuse information, especially among national CERTs.
Several approaches for interpretation of collected information and creation of situational awareness exist [32]. In principle, a manual, semi-automated or completely automated approach is feasible. In the latter case, very precise data are essential, should such data not be available, it is valuable to have human experts carry out assessments. Rulebased evaluation engines can aid with the evaluation of a situation. For instance, Cimino et al. [33] describe an adaptive rule-based approach for managing situation-awareness. They detail a generic software architecture for generating situation awareness and propose the use of fuzzy context ontology in order to generate a knowledge base. In addition to approaches that have been described in the scientific literature, there are a number of tools that can be helpful in the automated interpretation of a situation. Focusing on IT security incidents, the semi-automated framework YAFRA 4 aims to extract IOCs from reports and augment them with additional information from external sources. Cortex 5 an observable analysis and active response engine, which offers the possibility to analyze certain chunks of information (called observables, i.e., IP addresses, files, domain names, etc.). For example, Cortex allows to automatically check files or their hashes against online antivirus detection engines. It is also part of the Hive Project 6 , an open-source platform with the focus on incident response. Taranis NG 7 is another open-source platform, but has a broader focus on OSINT gathering and analysis. It aims to create structured report from unstructured news items. A highly specialized tool is Maltego 8 , which is used for opensource intelligence and forensics. Maltego aims to support an analyst with generating a visual representation of a wide variety of OSINT information blocks.
CCOPs can be represented in various ways, depending on the requirements. CCOP systems should provide tailored and timely information [4]. The representation can range from fully interactive web-based frameworks, over highly specialized 3D toolkits [34] to text-based reports. Existing visual analytics tools can be used in order to represent a CCOP.
Here, several open-source tools exist. Grafana 9 is a popular analytics and interactive visualization tool that may be used to visualize metrics, generate reports, or even generate alarms. Another example is Apache Superset 10 , which is concerned with the visualization massive amounts of data. The previously mentioned Taranis NG can generate reports via various presenter modules, such as PDF, HTML or simple text output.

Methodology for a cyber common operating picture
In this section, we define the requirements schema to guide the development of cyber common operating pictures (CCOPs), as well as an abstract model to develop a CCOP. We further define metrics that constitute the basis of CCOPs, i.e., the data that are visualized and interpreted in a CCOP, and roughly explain how data are ingested into the underlying data model in the first place.

CCOP requirements schema
CCOPs can be quite diverse, and depending on their application context, specific requirements arise along five dimensions, specifically if they are used to flexibly create cyber security reports: Scope and level: The scope and the target level of the situation reports shape the structure of the CCOP development process, as these strongly influence the selection of the information sources. At the operational level, for example, significantly more technical data sources and more detailed information are required than at the strategic level. Cyber situation centers aim to create strategic situation reports, which is why significantly more external context data may have to be included. Contextual data includes political, economic, and demographic information. For instance, it might be helpful to understand the background of an APT [30] or a wave of cyber industrial espionage and thus to create an early warning for potential victims. The aim is to be able to prepare long-term and medium-term decisions, therefore the essential relationships and possible forecasts must be prepared without detailed technical explanations.
Frequency and interval: CCOPs must be created periodically (daily, weekly, monthly). They must be available on an ad hoc basis in case of incidents, which implies a high degree of flexibility of the CCOP process. It must be noted, however, that the ad-hoc creation of CCOPs can lead to loss of information. Ad hoc reports (snapshots) can only incorporate the information available at one point in time into the construction process. Therefore, the completeness and thus ultimately also the correctness (at the time of completion) is deliberately neglected here. In case of ongoing attacks, cyber campaigns and possibly APTs, there may be greater deviations between the snapshots and the actual status, as the information basis and thus also the assessment of the situation can change quickly.
Output: Output consist of the results of the CCOP process, including warnings, incident reports or stakeholderspecific reports. The recipient of the output strongly influences the content as well as the level of detail of the description of the situation assessment. However, a compre-hensive description of the variety of display options is beyond the scope of this paper.
Sources of information: Particular consideration must be given to the diversity of the information sources in the process of creating a CCOP, as the quality of the sources has a strong impact on the assessment of the situation. A distinction can primarily be made between internal and external sources of information. Internal information sources include all information about an organization's own systems and services (e.g., relevant tickets and the extent to which the internal processes and capabilities are affected) and own information channels about new threats (e.g., intelligence information). The external information includes purchased cyber threat information channels, information from partner organizations and OSINT sources [30]. Information sources with appropriate degree of abstraction directly contribute to a situation report, while technical information sources are usually first prepared to meet a suitable form (e.g., the purchased threat intelligence sources often offer technical indicators such as IP addresses, FQDN, hashes, mutexes and the like. These data are compared with own log data and analyzed. The results can then be incorporated into situation reports in a suitable form).
Visualization: A CCOP must be able to visualize the current situation, historical situations, chronological events, the course of communication and the national cyber situation. The relevant information can be visualized in many different ways, e.g., as diagrams (bars, pies), trend lines, traffic light systems, tables with percentages or with the help of geographic maps.

CCOP process
After the definition of the requirements for a CCOP development process, we present a generalized model in Fig. 1. This model foresees different types of data and information "flowing" into the CCOP development process from a multitude of sources and via various transfer paths. Besides data collection, in the middle part, data aggregation and interpretation takes place and effectively "transforms" data into actionable knowledge. Eventually, the output process depicts this knowledge in an effective way to support decision making.
In CSCs, situation reports are created with different methods depending on their purpose, e.g., a report that supports the assessment of a current incident differs from a report that forecasts the impact of a recently discovered vulnerability.
Input: The input block includes all types of sources and feasible transmission paths to integrate them into the cyber situation center's CCOP development. Based on recent research projects [30], common sources that contribute to CCOPs are for instance incident tickets, asset information, CCOP development process: The general CCOP development process consists of four main steps: the data collection process (this also includes the storage of all relevant information, messages and historical reports), the aggregation process, the interpretation process and the output process. The individual steps are highly dependent on the application context and investigated in more detail in the case study in Sect. 5. The process of creating a situation report begins with the collection of data from a wide variety of information sources. This process runs permanently, as relevant information about cyber threats and attacks can arise at any time. With regard to the data collection process, a distinction can be made between automated and manual information retrieval. In the aggregation process, the data and information are assigned to predefined information classes. The interpretation process identifies, analyzes and evaluates the information relevant to the development process.
Output: The last step is the output process, where reports are generated for different stakeholders and needs, as well as periods of time and in different forms. The current security status can, for example, be displayed in the form of reports, statistics or dashboards at daily, weekly or monthly intervals.

CCOP metrics
In order to assess the overall cyber security situation, the state of each individual service is taken into account. A number of metrics are available for this purpose, which are explained in Table 1 for characterizing the overall cyber security situation within an organization, and in Table 2 to reflect the security status of a specific service. Notice, the last column shows for which type of reports these figures were considered in the case study in Sect. 5.
The metrics for the overall cyber security situation (Table 1) are structured into Cyber situation general, including various types of internal and external reports that reflect the current situation with respect to incidents; risk management, including the knowledge of risks, prevented risks and the effectiveness of risk management; and information security management figures that report on the implemented controls to mitigate risks.
On the other hand, the metrics for specific services ( Table 2) report information on the service status in general, including metrics such as uptime, data volume, costs etc.; vulnerability management, including number of found or fixed vulnerabilities and average times to discover or patch them; and incident handling, reporting on the number of incidents detected or handled, and average times and costs to resolve them. 11 These information, the general view on the infrastructure and the service-centric view, provide valuable insights to make justified decisions on the usage and maintenance of services. For instance, cost-intensive services with re-occurring vulnerabilities or frequently targeted by adversaries need to be protected differently or are used under more stringent constraints, compared to services not having these issues.
Both, Tables 1 and 2, include suggestions for evaluation intervals of the single metrics. While we suggest to calculate the value of most metrics on a regular basis, e.g., daily  or weekly-depending on the reporting period-for others it makes more sense to evaluate their difference to the previous period. This way, the change, e.g., a growth rate, can be calculated, which might be more expressive. For instance, the presence of remediation costs of 50,000 Euros might be less alarming than the increase in remediation costs by 50% since the last reporting period.

CCOP data ingestion
An essential part of creating CCOPs is getting the relevant data to aggregate and interpret into a common data model in the first place. We identified the classes given in Table 3 as viable data sources for CCOPs, which are commonly available as reports at different abstraction levels.

A common data model for CCOPs
This section focuses on the design and implementation aspects of a common data model that holds all the relevant information for the flexible creation of cyber security reports.
We also look at the processes in which this data model can be embedded to support the work of a CSC in a meaningful way.

Design patterns and requirements
In order to define the requirements for a common data model for CCOP, it is essential to understand the processes and working methods of a CSC. As noted in subsection 3.2, the process of generating a CCOP, and therefore the task of a CSC, is to interpret the information, some of which can be technical, from a wide variety of sources and to break it down to the essentials in order to inform the command level and facilitate decision making. In most cases, the CSC relies in this regard on the technical analysis and reports from other specialized organizational units, like incident handling teams or vulnerability handling teams. However, information from other sources such as websites, newsletters, OSINT reports or notifications can also be incorporated into the CCOP. Therefore, the tasks performed by a CSC differ from those of other units such as information security operations centers (SOC), whose tasks are located at a lower, more technical level. This places event reports in a central position in the CCOP process and requires them to be mapped into a generalized data model, which includes a mapping of typical internal and external messages and security-relevant incidents. The model must therefore also support an efficient analysis of the information available in the event reports. The model must further be able to represent the quality of an information source and allow the relevancy of information to be specified. Efficient analysis also relies on finding appro-priate information, hence it must be possible to make search queries efficient by keywording.
Since some of the original information must be used for analysis, the model must also have a way of making raw data available. These can be, for example, e-mail attachments or scrapped web pages. This also allows for better traceability, as it cannot be assumed that all sources of information that have led to a certain situation assessment will exist permanently. To further improve traceability, the model must support versioning and linking of references to all information sources (both reports and data sources).
As different data sources need to be integrated, the model should be extensible and not locked into any particular technologies. To aid interoperability and adoption, the data model should reference existing standards and conventions, for example certain vocabularies and terms.

The CCOP data model
The data model consists of various data-types, categorization inventories or vocabularies (modelled as enumerations) and relationships. Figure 2 provides a full overview of the data model in the form of an UML Class Diagram. In the following, these different elements of the data model will be discussed in detail.
The Report is the abstract base class for all events or messages that can be relevant for a CCOP. It contains information about the author, source, information rating and references. It can also feature a list of attachments. The CCOP data model defines a set of specialized reports which are relevant for the processes within a CSC: The EventReport is used to describe any type of security event. In addition to data fields for time classification, this report offers the possibility to define the incident type and the affected status. IncidentReport is derived from the EventReport and describes an security incident. It enables the mapping of the affected status of concrete assets within the own organization and the specification of a threat actor. According to NIST [35], an incident is an "occurrence that actually or potentially jeopardizes the confidentiality, integrity, or availability of an information system or the information the system processes, stores, or transmits or that constitutes a violation or imminent threat of violation of security policies, security procedures, or acceptable use policies." The CampaignReport describes an attack campaign. A campaign comprises a series of malicious activities or attacks (sometimes referred to as waves) carried out over a period of time against a specific group of targets. The description contains the target (desired effects), first and last sighting and, if known, the threat actor. The AuditReport represents a penetration test report. It contains one or more targets that were considered in the audit, the textual report itself, classification of the "depth" of the audits, as well informa- tion what was audited. It also describes the vulnerabilities (or findings) that have been identified within the own organization. The term vulnerabilities covers all inadequacies or defects in the requirements, design or implementation of the computational logic (e.g., code) in software and possibly also hardware components (e.g., firmware).
A Summary can combine one or more reports of different types. It features a textual summary for the reports contained. It also contains a risk level determined by experts, related to the own organization. Several versions of summaries can be stored via the History. The History can be used to implement versioning. This is done via a HistoryEntry, consisting of a mandatory time stamp (representing the modification date of the entry) and the corresponding Summary.
An Attachment is a representation of an arbitrary information object that is attached to a report. A specialization of the Attachment is the Indicator, which is an representation of an Indicator of Compromise (IoC). An IoC in IT forensics is an artifact that has a high probability of indicating unauthorized access to a computer [36]. It contains further information of the validity time span and the type of the IoC.
A DataSource contains all the meta-information of an data source. This includes the source name and type, keywords, a rating and if possible a DSCheckPoint. A DSCheckPoint allows to determine when a concrete information source was last interacted with, for example, when information was last imported from an RSS feed or when data was last processed from a log management system. As an example, the data model defines two typical checkpoints (RSS and Website) for data sources. However, it is expected that adopters will extend this list with their own data sources.
The data model defines several support data types, which are mainly used for the mapping of information. The StringTuple is the most basic data type and serves as an helper construct for various other data types. The AuditStatusMapping allows an assignment of a service, system or asset to one or more findings, as well as the audit status. If there are findings, they are listed individually as a finding and assigned a status. If a vulnerability has already been closed during the audit, the status "Closed" is set. A Finding describes a vulnerability found by an audit (or penetration test). It can also include additional infor- mation, such as a CVE number 12 or a CVSS rating 13 . The InformationRating allows to store reliability and relevance of an information and the SourceRating allows the storage of reliability of an information source, both in reference to an internal organizational unit.
The data model furthermore defines several types of entities (i.e., natural persons, organizations and attackers). An Entity can be a natural person, an organization or a threat actor. Entities themselves thus have the task of creating a basis for various forms of entities in the data model. Since they can also be a source of information, they are also a realization of a DataSource. Natural persons are mapped to Person and organizations to Organization. Both can have a Contact attribute, which encapsulates all contact information known. The ThreatActor is the realization of an entity for threat actors. Threat actors are individuals, groups or organizations that are believed to act with malicious The categorization inventories or vocabularies in the CCOP data model allow a precise classification of events, entities and processes. For instance, the AuditStatus allows the assessment of vulnerabilities or issues found during an audit the IncidentStatus allows to define the status of a security incident and the IncidentType allows specifying the type of a security incident and is based on IncidentCategoryVocab-1.0 14 . IncidentSeve rity allows for specifying the severity of an incident, based on the Cyber Incident Severity Schema [37]. Relevance allows the assessment of the relevance, in relation to the own organization, of a certain piece of information. SecAuditClass allows an assessment of the "depth" of the security audit. The selection of the appropriate value is made in accordance with Table 4. AuditTargetType describes the type of infrastructure (service, system or asset) to which a certain finding relates to. RiskLevel refers to the risk assessment of experts regarding the information from one or more reports for their own organization. The four levels shown here are intended as a proposal and are based on the AS/NZS 4360 standard; however, other methods are available [38].
Some categorization inventories are directly taken from existing specifications, for example IndicatorType is based on the indicator-type-ov and ThreatActor Sophistication on threat-actor-sophistica tion-ov of STIX 2.0 15 . InformationReliability and SourceReliability both use the categorization scheme from FIRST.org [39], which is again based on which is based on the US Army Field Manual 2-22.3 [29]. 14   ThreatActorType, which allows the type of a threat actor to be specified, is based on the perpetrator groups defined by the Canadian Centre for Cyber Security [40]. Options for extending this list are provided for example by Sailio et al. [41]. OrgType is to be seen as a first draft for a list of relevant organization types in the military context. VulnerabilityStatus allows the handling status of a vulnerability to be set.

Application of the data model
In order to use the data model presented here effectively, it is necessary to embed it in suitable processes. In this section, a generic process for determining the overall security status (i.e., the CCOP), which is nevertheless typical for a CSC, will be presented. Figure 3 depicts this overall process for generating the CCOP. Based on the established and wellproven approach found in the Austrian MoD, this generalized process for generating a CCOP has been derived. Here, two major views are relevant for the command level: an "external" view, which encompasses the general cyber situation worldwide, with a certain focus on the national level. In addition to this overarching cyber situation, an "internal" overview of all security-relevant processes, potential issues and risks and in the own organization must be created. Therefore, creating a CCOP consists of two main activities that can be carried out in parallel: (1) Determining the general cyber situation; and (2) determining the security status of the own organization via internal reports from specialized departments and other organizational elements. For this purpose, it is necessary for the CSC to be able to make use of different information sources. A list of the most common sources is provided by Table 3.
As shown in Figure 4, the general cyber situation is mostly deducted from different information sources. This includes public sources, such as news and media articles or OSINT feeds, and more private or restricted sources, such as reports from partner organizations and reports from internal departments or organization units. The process begins with information gathering from these different sources (see Report). If a source is new, it needs to be evaluated first (see also SourceRating in the CCOP DM). Internal doc-  Table 1 for an exemplary selection of metrics for these two fields) are an important part of the general overview of the security situation. Reports and statistics from incident handling and vulnerability management, as well as general service status information should also be included the CCOP. Table 2 provides a selection of metrics that give an indication of the security status of one' s own services. Information from external sources can, for example, be included in your own information system as an EventReport or CampaignReport. Information will also need to be evaluated (see InformationRating). This step is crucial for the final risk assessment of the information at hand. For instance, if new vulnerabilities become known, it must be checked whether they can also affect the own infrastructure; if new threat actors appear, it must be clarified whether one's own organization fits into their target scheme; if new TTPs become known, it must be clarified whether they have already been taken into account in the security controls; and also in the case of general events in the cyber security landscape, it must be investigated whether these are relevant for a situation picture for informing the command level. The information gathered will then be evaluated for potential risk to the own organization. This can be a manual task, but also machine learning techniques can be used. In some cases, consultation with other organizational units may be needed. This risk analysis depends on the needs and characteristics of the own organization. However, it is advisable to rely on established procedures, such as ISO 27005 [42] or NIST SP 800-30 [43]. Depending on the information situation or the required scope of the CCOP, it might be sufficient to only consider consequences and likelihood of a security risks, however considering asset value, threat, and vulnerability can significantly increase the informative value of this assessment. A statistical evaluation of certain security-relevant metrics (see Tables 1 and 2) can also be included in the risk assessment. At the end of this process, a report will summarize the results of the evaluation).
A more specialized workflow is shown in Fig. 5. In this case, we assume that our own organization provides different services (e.g., email or telephone) that use one or more systems (e.g., mail server, DNS server, VoIP server, etc.), which in turn consist of different assets (such as the physical hardware or VoIP phones). The infrastructure should be mapped in service, Asset and CMDB databases to support targeted evaluation. In order to gain an overview of the security situation of one's own infrastructure, all available relevant internal reports (see also IncidentReport and AuditReport) must be evaluated. The relevant reports can be selected via the keywords and the stored meta information (e.g., timeliness/date, scope, creator, etc.). If certain systems or assets are not linked to a service, reports on these must be evaluated separately as shown in Fig. 5. Since internal reports prepared by qualified experts are used here, their evaluations can be used. In the case of incidents, an assessment of the severity of the incident should already have been carried out. If the incident is still ongoing and has not yet been fully dealt with, it is necessary to consult with the incident handling team. The same applies to security audits; an assessment should be available for each infrastructure element examined. If a CVE number or a CVSS score was specified in the finding, these can also be used for the assessment [44], [45]. At the end of

Case study with the Austrian MoD
In the end of 2021, a case study together with the Austrian MoD was carried out, which carefully followed the design methodology described in this paper.

Use case context
In this case study, CCOPs are used to capture the current cyber security situation and serve as the basis to flexibly create cyber security reports for specific stakeholder. With respect to the CCOP requirements defined in Sect. 3, we defined the requirements summarized in Table 5. As shown in Sect. 4, the CSC relies heavily on reports and information generated by other departments, experts and sub-organizations. For this case study, the following entities within the Austrian MoD had to be taken into account: Target level Strategic/operational The CCOPs aim at supporting decision making and the preparation of forecasts for longand medium-term decisions. Therefore, the information sources with the appropriate abstraction levels must be used in the creation process.

Duration
Periodic, ad hoc In the military cyber situation center, both periodic (daily, weekly, monthly) and event-driven (immediate) CCOP building is desired. This requires a flexible process.

Output
Alerts, incident reports, reports tailored to the needs of the user. Warnings for awareness building for users. The findings of the situation assessment are summarized in the form of cyber alerts, incident reports, and situation assessment reports. The CCOP building processes must be able to provide the information in formats appropriate for the user.

Sources
Internal and classified external sources, and OSINT The diversity of information sources must be particularly well managed in the CCOP building process, as the quality of the sources strongly shapes the situation assessment.

Visualization
Current and historic situation, chronological events, communication history, overall cyber situation The CCOP should be capable of recording and reproducing the current situation (including the situation of the whole country) as well as the past situation, events and communication.
-The cyber situation center (CSC) is responsible for generating situational reports for the purpose of providing information to the command level. These reports can be generated from different occasions. Basically, a distinction is made here between two types of reports. Regular, periodic reports (e.g., daily reports, weekly reports or quarterly reports) cover a defined period of time and contain all relevant information necessary for decision makers. In contrast, there are ad hoc reports that focus on a specific topic, such as an incident. The information sources that are potentially relevant for a situation report are shown in Table 3. -The Point of Contact (PoC) is a defined point of contact for external partner organizations. Different partner organizations may have different PoCs, which may also differ according to the concern. -The Command Level is above the other actors involved in the command hierarchy and has the authority to make decisions. These decisions are made, among other factors, based on the reports of the CSC. -The Incident Handling (IH) is tasked with dealing with security incidents. The focus of the IH is of a technical nature, the IH thus performs the tasks of an SOC. -The Vulnerability Management (VM) is responsible for dealing with technical vulnerabilities in MoD services, systems, and assets. The VM's duties also include performing technical audits, such as penetration testing. -A Service Owner is the operator of a specific IT service within the MoD. These services can be specifically assigned to a department or unit. The service owner here is also responsible for asset management and infrastructure monitoring.

CCOP process implementation at the Austrian MoD
In order to validate the CCOP generation processes and the corresponding data model, a number of exemplary use cases were defined. One objective of these use cases is to embed the processes described in Sect. 4.3 into an overarching process. One of these use cases is modeled as a business process in Fig. 6 and will be discussed in this section. This use case includes a wide array of information sources and means of technical communication. The situational picture to be drawn here has to be included in an ad hoc situation report of the CSC for the short-term information of the command level.
In the fictitious scenario presented here, an APT 16 has been detected in a partner organization (PO). Based on traces left behind by the attackers, the PO's security experts conclude that the attack must have been carried out by state- 16 The description of the APT is adapted from the real-life APT40, which became known in 2019. [46] sponsored hackers from the country of Ostarrichi. The initial infection took place most likely in January 2020. The PO then sends an encrypted mail to a defined point of contact (PoC) at Thursday, November 11, 2021, which contains the following information: -Tactics, Techniques, and Procedures (TTPs) [47]: The APT was most likely introduced via spear-phishing mails with malicious attachments or with links to malicious files on cloud storage drives. Attackers leveraged exploits in Microsoft Office and .NET products. The APT relied on web shells for an initial foothold. The APT targeted then VPN and remote desktop credentials to establish the foothold. To evade detection, the APT used well-known web resources as Google Drive, Microsoft OneDrive or Apple iCloud for file exchange. This report triggers a series of internal activities. In the first step, the PoC identifies Incident Handling (IH) as the internal organizational unit that must be informed first. If an incident should also occur in the own organization, the IH can react quickly and, if necessary, initiate mitigation measures. The mail with the report is therefore forwarded to IH by the PoC.
The IH begins with a relevance check of the report. The IH clarifies here, for example, on the basis of the aforementioned security vulnerabilities, whether its own systems can be affected at all. In addition, the assessment includes whether the own organization has structural and functional similarities to the PO, and can thus be a potential target of the hacker group. The IH may also use the Information Source Catalogue to access information on the reliability of the partner organization (see DataSource with SourceRating); if no assessment is available, this can be carried out for the first time by the IH. If the information contained in the report Fig. 6 Example use case: security incident at an external partner organization is deemed relevant, i.e., if the own organization is possibly also affected, the report is included together with a Summary (including a first risk assessment, see RiskLevel in Fig. 2) in the work-flow management system (WMS) and thus the CSC is informed automatically. In this case, the IH concludes that the own organization falls within the profile of the attackers and that its own systems potentially could have been affected by the vulnerabilities.
The CSC now starts to collect all available information in order to inform the command level by means of a CCOP. In this specific scenario, two different entities are being contacted, on the one hand the internal Vulnerability Management (VM), and on the other hand the national CERT, which is an external organization. The VM is contacted via internal chat to find out the current status of the company's own services and systems. A reference to the summary with the original report, including the list of vulnerabilities in WMS, is provided to the VM as well. The VM begins to compile a report (see AuditReport), which concludes that currently, there are no known open vulnerabilities in the infrastructure. However, the report notes that the vulnerabilities listed in the original warning from the PO were once present, with CVE-1337 mitigated on 2021-08-09 and patched on 2021-09-01, as well as CVE-0815 and CVE-4711 both mitigated on 2021-09-05 and patched on 2021-10-01.
The CERT is contacted via encrypted mail to find out whether similar incidents are known or whether there is a major wave of attacks (i.e., campaigns). The CSC can obtain this contact information from the CERT profile backed by the data model (Organization with a ContactPoint). The CERT reports back that there have been several suspicious activities which are connected with Ostarrichi in the last months. For instance, critical infrastructure operators have reported that there have been probing attacks from the Ostarrichi IP address range. The CSC receives this information and records it as a CampaignReport. The information of the adversary from this report is integrated into the ThreatActor entity which is assigned to the original APT report from the PO.
In parallel to these processes, which have been kicked off by the CSC, IH is focusing on the incident handling process. Here, the IH begins to set up initial mitigation measures, which include entering IoCs into detection systems, creating new SIEM rules, analyzing logs, and so on. The IH also starts an investigation to find traces of the TTPs, respectively, IoCs, on its own infrastructure. The IH detects that there are suspicious attachments on the mail server of the Public Rela-tions Department. As a result, mail service and operation of this department's workstations will be temporarily suspended pending a comprehensive analysis. This is carried out in coordination with the service owners, here the operators of the mail server and the user client network. An ongoing APT can neither be confirmed nor ruled out at this point. The results are recorded in an IncidentReport. This report is then attached to a new instance of the original Summary, together with the original report and an updated risk rating.
Once the various information and reports have reached the CSC, it can begin to create a CCOP that reflects the current status. As quick reaction is required in this ad hoc cyber situational picture, this is done in the process detailed in Fig. 3, with a focus on the sub-process displayed in Fig. 4. The evaluation of the historical vulnerability reports has been carried out by the VM. The initial report of the PO (already assessed by IH), the incident report of IH, the vulnerability report of VM and the campaign report of CERT are selected by the CSC as the basis for the situation report to be prepared. Since there are already risk assessments by the IH and VM, they can be incorporated into the final situation assessment in an automated process. This also applies to the existing relevance assessments for information and reliability evaluation for information sources. Clear guidelines for assessment are absolutely essential, but are not the focus of this work. Ultimately, however, a final sanity check must be made by security experts before the report is published. In the present scenario, IH and VM experts should also be consulted for a final assessment. No information matching the APT was identified in the relevant OSINT feeds and specialized media. Based on the risk ratings from IH and VM, the CSC concludes that the own organization is possibly at risk.
Finally, the results are summarized in a situational report for the command level (see Fig. 7 for a truncated version). The ad hoc report contains a brief summary of the current situation, including an assessment of the situation by the CSC, partly based on the assessments given in the IH and VM reports. In addition to the current incident, the situational picture shown here contains other exemplary elements. The upper part briefly summarizes the assessments on various topics (in addition to ICT services abroad and domestic, the cyber security situation, which is the actual subject of the report). The metrics for services (see Table 2) are aggregate into combined indicators for (in this scenario) three distinct service classes (mission critical, essential, and other) for the different departments of the own organization. For instance, a high number of incidents or a high number of unmitigated incidents can lead to a warning ("yellow traffic light"), in severe cases to an alert ("red traffic light"). The threshold values that lead to this combined indicator going to another warning level must be defined in cooperation with service operators, command level, CSC, IH, VM and possibly other involved stakeholders (e.g., user representatives).
The statistical data for effectiveness of risk management and information security management activities (see Table 1) are consolidated into a graph, which represents their respective trend for the various regular reporting periods. As with the services, the idea here is to calculate a combined indicator value from various metrics, which will provide the command level with an overview of the trend in the effectiveness of these activities. Internal and external reports from this table are also integrated into the situation picture. These information blocks can be automatically generated and integrated into any situational awareness report-be it ad hoc or periodic. The visualization shown here is to be seen as a draft of how these types of information can be prepared into a (static) situational report. Using dashboards, such as the open source tools Grafana 18 , Chronograf 19 or Kibana 20 , this data can also be visualized interactively and in virtually any desired form.

Proof of concept application
The proof of concept application is a browser-based application. The implementation covers the defined use cases and uses the designed CCOP data model to gather, modify, and store data. Hereby, actually two databases are used. The first database contains persistent data, whereby persistent data refers to the original data sources and stored common operational pictures. The second database is used to work on a near real-time operational picture that represents the current status. All modification made to data and all additional information added by a user will first be cached in this second database and only be persisted once the user decided to save a snapshot in the form of a report whereby the report could be a daily, weekly or monthly report, as well as an ad hoc report.
The goal with the proof-of concept application is to demonstrate that our solution makes the creation of cyber security reports more efficient. In order to achieve this goal, the application is split into several views, whereby the three major ones are: With these three views, the proof of concept application follows the CCOP development process of data collection, aggregation, interpretation and the output process outlined in Fig. 1. The data collection however is partly performed outside the proof of concept application by saving data into our database or creating events in other applications. The proof of concept application uses this data to display relevant events and messages/notifications for the CCOP development process (cf. Fig. 8).
Having the discussed use cases in mind the proof-of concept application follows a logical sequence of steps supporting the various tasks of the CCOP development process. As such, the user can view events and messages by selecting an entry from the list, mark events that are of relevance for a report, combine them or enhance them with new information as part of the data collection and aggregation step. Hereby, it is possible to select for which kind of report the event is relevant, for a daily, weekly, monthly, or ad-hoc report. The attachment of other relevant information to a created or edited event, be it other messages or documents, is also possible. With the ability to further add and describe countermeasures also the interpretation step is supported.
Switching to the "Lagebild" view the user gets information about the current situation. As the proof of concept application focuses mainly on the creator role of a CCOP the last entry in the time line of the top of the page represents the currently worked on status. The other entries represent CCOPs that have already been persisted and saved to the database. The overall layout although is the same, whether the user is a creator or recipient of a report. On the left side, always the summary is shown, while on the right side details to a selected entry in the CCOP will be shown. The recipients can in general not make any further modifications.
A role management mechanism defines access rights so that only roles defined with appropriate rights can perform modifications to some information already in this view. This page is the main page related to the interpretation step of the CCOP development process since here the information relevant for a report is marked, enhanced, and measures are added to an event. Once a user with appropriate rights decided to build a specific report by selecting the kind of report required and pressing the Create-button on the lower left side of the page a snapshot of the currently shown (filtered) situation is taken, and information marked as relevant for the chosen report type are copied and a new report is prepared as shown in Fig. 9.
Roles with appropriate rights can still make modification to this report. It also still uses the temporary database to store information and changes. Only after pressing the Create-button this report is persisted and saved into the other database for later viewing. By pressing this button, the report also becomes available to other recipients, finalizing the output process. Additionally, the proof of concept application offers the possibility to view and search for persisted reports in the database. The "Suche in Lagebildern" allows to define a time frame. Reports within this time frame are displayed in a list and can be further narrowed down by additional search terms (see Fig. 10).
Details of the reports will then be displayed on the right side of the page once an entry is selected. The overall layout of the HMI was defined together with the stakeholders at the Austrian MoD to cover the relevant information and to be flexible to support the needs of various stakeholders in the information chain, the creator as well as various kinds of recipients which all have slightly other information rights and needs. Dependent on the rights and needs of the recipient of the report some of the text fields and information may be filtered out and thus is not visible to them. The screen shots used in this section however are taken from the report creator role, thus all information and all fields are visible.

Discussion and evaluation
The proof of concept application was validated by means of a virtual presentation, in which not only the prototype was presented, but also the main use (creating an overview, creating a situation picture) was demonstrated. Meanwhile, questions and suggestions were welcome. After the presentation, questions were asked in a focus group, which led to further discussions. Finally, the well-established System Usability Scale (SUS) questionnaire [48] was distributed to evaluate the usability of the implemented application. The main scenario was also tested with future creators of CCOPs in order to receive feedback on the information hierarchy, the presentation and the process, and to be able to draw qualitative conclusions about usability.

Test process
Basically, the test process was composed of three phases: -Phase I: Pre-test-In a pre-test, it was the task of a test participant to carry out the planned scenarios with the help of the PoC. Important usability problems were

Test procedures
All phases were carried out according to well-defined procedures that were designed as follows: 1. Introduction to the system: We provided a brief introduction to the topic of the project and the PoC, as well as an overview of the main activity of the concerned CCOP end user. 2. Execution of the main scenario: The individual steps required to create a CCOP were shown on the prototype. 3. Focus group: Based on prepared questions, the practicability, comprehensibility and possibilities for improvement of the prototype were discussed in a focus group. 4. Distribution of the SUS: The SUS questionnaire was electronically distributed to all participants and feedback collected.

Validation
Execution of the main scenario. The participants carried out the following steps, in order to evaluate the degree of support that the proof of concept application provides for the creation and assessment of CCOPs: -Create an overview of new messages (cf. Fig. 8) -Verification of the current events in the "live" situation picture (cf. Fig. 9) -Open the current situation picture -Addition of general information -Verification of current events -Set current status and trend -Review of the "live" situation picture -Creation of a daily situation picture (snapshot of the "live" situation picture) (cf. Fig. 10) -Selection of the template (daily situation picture) -Verification and supplementation of the content -Enter the planned release -Approval of the situation picture Focus group. After the demonstration, a set of questions were asked and discussed in the group. These questions were the basis to discuss any open issues, collect individual feedback, but also to prepare the participants for the final questionnaire, designed according to the SUS methodology. The following questions have been asked: -Does the proof of concept support you in completing your tasks without burdening you unnecessarily?
-Are all functions available? -Is all information available? -How could work-flows be optimized?
-Is the proof of concept sufficiently understandable? -What could still be improved?
Calculation of the SUS. The SUS is a way of simply evaluating the usability of a product or system. It consists of ten usability statements that are rated on a scale from 1 (totally agree) to 5 (totally disagree). The SUS contains five positive and five negative statements on the usability of the system to be evaluated. Based on this evaluation of these statements, a usability value is calculated, which can then also be used to compare with other products. The final score, calculate among all participants, was 68.1 as also illustrated and interpreted in Fig. 11.
Qualitative key statements. During the interviews hundreds of qualitative statements were collected that form the basis for further improvements. Besides quite detailed feedback regarding the user interface design and general usability issues, most feedback revolved around a few topics: -Usability improvement: The positioning of user interface elements, the font and the positioning of information elements should be investigated in more detail in course of a user experience evaluation. -Tool architecture: Possibilities for off-line usage (e.g., abroad missions) and isolated operations should further be investigated. -Transparency and auditability: It is important to understand how a CCOP was created, how the different informational elements were evaluated and a current situation assessed. If a human analyst performed some of the steps, it should be transparent who did the work.
-Automatism and interfacing: More interfaces for enrichment with further information, e.g., photos, geographic information and non-IT systems are essential to complete the picture. Automatically fetching this information, annotating and linking it to is further vital for swift CCOP creation. -Information labelling: Manual annotation, tagging and justifications, e.g., of derived and interpreted situations shall be stored, however a high degree of automation shall be achieved with as little human involvement as possible.
With regard to that processing of classified information, e.g., according to the traffic light protocol, must be enabled. -Going beyond state-of-the-art: Today's processes involve lots of manual effort, such as copying and pasting in standard tools (such as MS Office). Centrally collecting information, having them processed and interpreted by dedicated actors, and having them distributed along predefined lines, has tremendous potential to streamline and optimize existing processes.

Conclusion and future work
This paper presented an approach to create cyber situational awareness through collecting and interpreting messages, events and notifications from various organizational data sources. Current technologies that are utilized such as e-mail, ticketing systems, shared documents or CTI tools provide some functionality to exchange information; however, these are usually spread across different systems and services and therefore do not efficiently support the fast creation of cyber situational reports. Therefore, we proposed a new approach and introduced a common data model that frequently captures all the relevant information in a coherent manner that are used in the CCOP creation process. Two major views are relevant for the command level: an external view, which encompasses the general cyber situation world-wide, with a certain focus on the national level; and a second view which provides insights on the threat level of the own organization only. The investigation of typical metrics complemented these efforts. We analyzed the applicability of this approach by implementing a proof-of-concept application together with the Austrian ministry of defense that quickly enables the re-occurring or on-demand creation of reports (e.g., in case of incidents). A novelty of our solution is, that complex and diverse data elements, relevant for cyber security, can be collected over a multitude of interfaces and reviewed, interpreted and reported upon in an aggregated form at a central point. From there, adequate mitigation actions can be taken, relevant stakeholders informed, and the cyber situation tracked over time.
Future work includes the in-depth evaluation in the operational military environment, were future users will be challenged with realistic situations. This evaluation will focus on the applicability of the tool in the target environment on the one side (e.g., fulfillment of domain-specific requirements, etc.) and overall usability aspects on the other side. Regarding the further development of the tool, more automation for the evaluation of a situation based on data inputs using, for instance, rule-engines for reasoning, seems to be a promising field.
Funding Open access funding provided by AIT Austrian Institute of Technology GmbH This work has been funded by the Austrian defense research programme FORTE of the Federal Ministry of Agriculture, Regions and Tourism (BMLRT) in course of the project CADSP (873425).

Data availability
The datasets created during the usability evaluation are not publicly available due to their military background.

Conflict of interest
The authors declare that they have no conflict of interest.
Ethical approval This article does not contain any studies with human participants or animals performed by any of the authors.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecomm ons.org/licenses/by/4.0/.