MARISA provides a data fusion toolkit for enhanced maritime surveillance — importantly, including data from the internet and social networks — to “improve information exchange, situational awareness, decision-making and reaction capabilities” through a solution that is capable of providing “mechanisms to get insights from any big data source, perform analysis of a variety of data based on geographical and spatial representation, use techniques to search for typical and new patterns that identify possible connections between events, explore predictive analysis models to represent the effect of relationships of observed object at sea” (video presentation).
Deliverable D2.7 also lists the project’s data sources: “Very large amounts of data, ranging from optical sensors, satellites, radar, AIS (Automatic Identification System, ed.), open and internal sources.”
Furthermore, the “MARISA toolkit provides new means for the exploitation of the bulky information silos data, leveraging on the fusion of heterogeneous sector data and taking benefit of a seamless semantic interoperability with the existing legacy solutions available across Europe, with a particular reference to CISE (the Common Information Sharing Environment for the maritime domain, ed.) context, whose data model and relevant use cases will be adopted, in order to demonstrate the MARISA capability to provide value added information (i.e. alarms, events, risks, anomalies, patterns, etc.) upon the use cases and scenarios already defined within previous projects (CoopP) thanks to the processing of input data coming from CISE nodes and / or from users legacy systems.”
The deliverable also speaks about “end-user benefits”, such as obtaining “a multinational maritime picture to be routinely used during illegal crossings from third countries to Europe” and an “early warnings of smuggling, illegal immigration, terrorism before they reach nation’s coastline.”
The project’s final report, D1.4, uses an interesting metaphor to justify the notion of “data fusion” adopted in MARISA, arguing that “the process of integrating multiple data sources to produce more consistent, accurate, and useful information than that provided by any individual data source (…) is analogous to the ongoing cognitive process used by humans to integrate data continually from their senses to make inferences about the external world.” Maritime surveillance is, in other words, conceived of as a complex, seamlessly interacting living organism that only works if all components are correctly integrated.
The MARISA toolkit consists of various technological innovations, including:
1) “new algorithms to detect ships in order to decrease false alarms in difficult sea conditions (low wind areas with small local pixel value variations and high wind areas with peaks created by the breaking waves pattern, both sources of false alarms for traditional CFAR (Constant False Alarm Rate detectors, ed.) and to better estimate ships parameters (e.g. heading, length, width);”
2) “Integration of OSINT (Open Source Intelligence, ed.) from the Open Source Global Database of Events (GDELT) identifying and classifying maritime safety and security events from multiple languages to improve the Maritime Situation Awareness;”
3) “use of VHR (Very High Resolution, ed.) optical data and SAR data, also open sources when available, with more accurate vessels parameters estimation;”
4) “integration of both satellite derived and ground based AIS data;”
5) “improvement of the situational awareness providing all available different vessels’ data including characteristics, images, videos, anomalies, alarms and threats;”
6) “track fusion algorithms designed to avoid data duplication on maritime picture through a data management strategy based on the innovative concept of M-track data entity. The M-track handles the full-time perspective of a vessel when it is monitored by several independent legacy systems along its track history.”
In a project video presentation it says that “Anomalous Behaviour Detection” happens automatically, and makes use of: 1) AI, 2) Machine Learning; 3) rule based methods; 4) probabilistic methods; 5) Satellite Images Processing, “integrating detailed forecast of weather and sea conditions.”
At the “decision-making” level, MARISA also claims to provide 1) predictive analysis, 2) threat assessment, and 3) mission planning.
MARISA was announced to be a part of “CISE related projects”, which included ANDROMEDA, ROBORDER, and RANGER, in a post by EMSA that was later removed.
The MAiDEN project was also mentioned in a post that was later removed.
Deliverable D1.7 adds: “The approach for the ethics work in LAUREA (the University of Applied Sciences in the Uusimaa Region, Finland, where MARISA’s ethical approach was discussed, ed.) has been developed during several EU project (including INAHCUS FP7 and H2020 projects WELIVE, MARISA, RANGER, ANDROMEDA, SHAPES). MARISA project was the first project where the ethical issues were systematically operationalised as ‘Ethical requirements’.” After MARISA, this work has been continued in the ANDROMEDA and in the SHAPES project: “Our plan is to develop the approach into a comprehensive model which can be published and developed upon by other researchers.”
The project’s Final Report, deliverable D1.4, further adds that “several European research projects have been already funded where all major providers of maritime surveillance solutions in Europe have collaborated between themselves as well as with the European authorities.”
A figure lists BLUEMASSMED, SeaBILLA, PERSEUS, CLOSEYE, and EUCISE2020. Importantly, the document underscores that “MARISA has carefully analysed the achievements of the previous investments and has built on their results.”
Use cases, operational scenarios, and operational trials are detailed in deliverable D2.7.
Use cases include the “monitoring of all events at sea in order to create conditions for decision making on interventions” and “request any information confirming the identification, position and activity of a vessel of interest.”
We also learn that five operational trials were planned, namely the
1) Northern Sea trial, “to validate a decision support tool in a relevant environment. The decision support tool fuses heterogeneous vessel information, detects risks and threats, and gives an advice on how to allocate the available resources for mitigation or interdiction;”
2) Iberian Sea trial, “to enhance the situational awareness of the end users (for example at the different National Coordination Centers) enabling the interchange of information among them through the MARISA toolkit (interestingly, “selected events” can be imported/exported through the dedicated EUROSUR (the European Border Surveillance system, ed.) gateway, thus ensuring a proper integration with EUROSUR”);
3) Strait of Bonifacio trial, to enhance “Traffic Safety through a better exchange of information and fusion of different data types, and the provision of added value information related to the identification of threats and response planning by combining OSINT data, radar tracks and optical data” (notably, tackling “irregular immigration” is included);
4) Ionian Sea trial, both to detect “possible abnormal behavior patterns of vessels used for human trafficking” and to obtain “better performing automatic tools for earlier detection of suspicious vessels, abnormal vessel behavior, and identification of threats even in high (summer) traffic density through contextual information from open systems (Social Networks, etc.) correlated with fused basic data types;”
5) Aegen Sea trial, “to demonstrate the end users situational awareness enhancement by using MARISA services to detect human trafficking activity in the Aegean sea.”
While the project’s website now redirects to a sketchy online casino, some updates concerning their results can be gleaned from MARISA’s Twitter feed: “To demonstrate the added value of #Maritime Situational Awareness (#MSA), the MARISA project successfully carried out MARISA operational Trials in 5 regions, resulting in improved MSA for participating Navies & Coast Guards.”
End users for the trials included the Italian and Portuguese navy, the Dutch coastguard, the Greek MOD, and the Spanish Guardia Civil.
According to the project’s final report (D1.4), the full results and an “exhaustive analysis” are included in deliverable D7.7, “MARISA Operational Trials Results Report and Lesson Learnt”, which was however not available to the public at the time of writing.
D1.7 states that MARISA is designed “to create technology-based solutions for the market”, before adding: “One particular challenge that relates to trials and pilots is that there typically exists a tension between the purpose of a trial (validating the features of a solution) and ethical requirements (which should already be validated). In other words, any beta version of a solution must itself fulfil the minimum ethical requirements defined – that an activity is taken as part of a test phase is no excuse to shirk ethical responsibilities. For MARISA this was a challenge especially when it came to data protection legislation.”
This applies particularly to the analysis of Twitter data: “The use of Twitter data during the second phase pilots raised concern among Data Protection Team. Although the information of the ID of the twitter message was anonymized, there possibility that the message itself would include personal data remained. Also the EU Ethical Review Report in 3/2019 identified the use of social media data as one of the challenges for MARISA. Based on the results of the DPIA and a separate literature analysis conducted by Dr. Jyri Rajamäki from Laurea, an evaluation was made with the conclusion that the use of Twitter data during the MARISA trials did not constitute a risk for the data subjects’ rights (see Appendix E). However, the use of Twitter data remains a challenge for MARISA especially when it comes to the solution’s future use after the project is over.”
“Cost-effective” technological solutions are deemed necessary to efficiently manage migration through enhanced coordination and data sharing among maritime surveillance authorities, as
the project coordinator explains and ultimately appears in all project deliverables. It is therefore not surprising that the broader CISE initiative has the same framing: “CISE will ensure effectiveness and cost-efficiency of maritime surveillance activities.”
Allegedly, “technology, architecture, user processes, future governance” and even “business models” have been defined “in compliance with ethical and societal requirements set by EU fundamental rights and the GDPR.”
Ethics-related documents however provide some interesting insights. MARISA’s and RANGER’s ethics assessment have been discussed in the paper “An Ethical Framework for Maritime Surveillance Technology Projects”: “Due to the fact that both EU law and various international conventions regarding e.g. human rights, the rights of refugees, and Search and Rescue activities impose obligations on states to help and protect those in need, the increased situational awareness enabled by the new technologies will also lead to an increased responsibility to act.”
About the Ethics Appraisal Scheme for HORIZON 2020 projects, it says: “The ethical guidelines of Horizon2020 focus heavily on traditional research integrity-related issues and as such offer limited guidance in terms of the ethical and societal sustainability of real time setting piloting and trials, or the final product/solution itself. This can pose challenges for developers.”
It would be important to “remember that the features of the developed solutions may vary according to the environment in which they will be implemented, which may have implications on ethical requirements on all layers of the solution. Both MARISA and RANGER, for instance, can be implemented as either stand-alone solutions or as part of the Common Information Sharing Environment (CISE)”.
The Conclusions warn against shallow ethical assessments that are structurally incapable of getting to the projects’ core moral issues that should be publicly discussed in democratic societies:
“It should be evident that ensuring the proper implementation of ethical requirements is essential for any project. In spite of this, ethical compliance has long been near synonymous with proper research ethics, other important dimensions having been left with more or less an anecdotal status. The problematic nature of such a narrow perspective is often particularly accentuated in cases where a project’s subject matter falls under a security topic. The RANGER project provides an illustrative example of this: when technological advancements lead to an increased surveillance capacity for authorities (in this case in the form of novel over-the-horizon radars), so do the moral and legal duties to act against ill will and to help those in distress; with great power comes great responsibility. Furthermore, the developed technology can fundamentally change practice and customs: the moral division of labour can be altered, a change that calls for holistic ethical considerations.”
D1.7 — the Societal Ethical Final Report — notes that the “MARISA Legal, Ethical and Societal Aspects” deliverable comprises a literature review on the ethical and legal background of the maritime surveillance domain, an analysis of the main challenges related to the use of MARISA, and a Societal Impact Assessment (SIA). A list of ethical requirements and a Code of Conduct were presented at the end of the document.
