EURMARS is emphatically described as “a ground-breaking vision” and a “paradigm shift in maritime security”.
This self-proclaimed surveillance and maritime awareness revolution is built on the idea of stacking multiple surveillance technologies and have them work together to provide the authorities with actionable information in real time. This information would be then operationalized according to a risk-based framework.
More precisely, the project’s website writes: “EURMARS’s ground-breaking vision is to expand the common risk assessment practices currently deployed by authorities to enable the development, deployment and evaluation of a secure multitasking surveillance platform that improves sensing capabilities for a wide range of security risks and threats in wider border areas by clustering high altitude platforms technology, satellite imagery, UxVs (uninhabited vehicles) and ground-based sensors into a novel joint surveillance capability. As part of the EURMARS framework, the various existing and future systems for maritime surveillance will be integrated, to allow for the collaborative operation and the provision of the sensing results to related authorities. The open architecture will build on the lessons learnt of previous initiatives, assimilate the knowledge of the stakeholders and their practice on CISE and other relevant systems, exploit the latest AI, risk assessment and visualization innovations, and undergo extensive technical and user acceptance tests and ethical and legal impact assessments.”
Concerning the alleged “paradigm shift”: “The EURMARS project, stands at the forefront of a paradigm shift in maritime security. The integration of cutting-edge technologies, AI-based systems, and collaborative frameworks signals a commitment to fortifying the EU’s borders against emerging threats. As the Coastal Ground and Low Altitude Sensing Systems take shape, the consortium moves closer to realizing a future where comprehensive surveillance ensures the safety and security of European waters. The visionary approach of EURMARS, is set to redefine the standards of maritime security in the years to come.”
The solutionist framing of this description — which essentially assumes that a seamless integration of all the latest technological developments is needed to provide a solution that ensures the safety of European waters — is manifest.
At a general level, EURMARS involved various types of unmanned vehicles (UxVs), AI, “Ground-breaking enhancements of sensor capabilities to detect critical objects of interest”, a “decision-making tool” (“Delivery of advanced decision-making support tool and enhanced user interface under a cyber secure platform”), “a flexible platform with verified easy-to-integrate potential for next generation platforms and systems.”, and benchmarks “for adopting advanced data fusion and AI-based analytics in maritime operations.”
Several details are however currently lacking, except for the components illustrated in a couple of articles published on the project site:
1) a first article, dated April 2024, details “Skyld Ltd’s Contribution: Coastal Ground and Low Altitude Sensing Systems”, and writes: “This module is designed to generate reliable geo-referenced detections and tracking of ships, small vessels, persons, and vehicles in real-time under challenging maritime conditions. The UAV Platform utilizes airborne camera systems triggered by abnormal events detected by other sensors, verifying and confirming events during patrols.”
Its “Technical Specifications” were also “unveiled”: “The intricate design of the Coastal Ground and Low Altitude Sensing Systems is a testament to Skyld Ltd’s commitment to innovation. The system incorporates:
– Camera Sub-Systems: Combining shortwave IR, UV, thermal, and RGB cameras with ROS2 software libraries for live/raw image processing.
– Vessel/Vehicle Classification Sub-System: Employs PyTorch for offline training on representative datasets, ensuring real-time classification using GPU technology
– Behaviour Analysis/Anomaly Detection Sub-System: Developed in Python, leveraging MQTT message broker for seamless integration with other components.”
2) a second article, dated October 2023, details the MuFASA (Multimodal Fusion Architecture for Sensor Applications) data fusion module: “The overall vision of the project is to develop a platform that will improve sensing capabilities for wider areas by integrating high altitude technology, satellite imagery and UxVs in addition to ground sensor platforms in order to prevent, detect and react to crime, including that crossing external borders, illegal border crossings and/or smuggling at the border regions of the EU and of the Schengen area. With this challenge a wide arsenal of sensors and external data sources is needed to withstand the complexity of the use-cases. Thus, a sophisticated data fusion approach within a modular architecture is essential”.
The article goes on to specify its “central role in combing homogeneous and heterogenous data” with the goal “to decrease false alarms by combining different data sources as well as increase measurement precision to metadata interpretation”.
This requires “different individual data fusion modules”, which “are being developed or further improved within EURMARS”.
Importantly, “Some of them already have shown great promise in previous projects (e.g FOLDOUT), such as the MuFASA (Multimodal Fusion Architecture for Sensor Applications) developed by AIT (Austrian Institute of Technology, ndr)”, which “will be further improved and developed” during EURMARS”.
According to the article, “MuFASA excels in reducing the overall false alarms produced by single sensor systems”, and “this is considered as one of the main impacts of the EURMARS system”.
Among the data fusion models that EURMARS is developing and improving we find “MuFASA” (cfr. above), which was initially developed within — among others (“e.g.”) — the FOLDOUT project.
Deliverable D6.1, which we only obtained after an access to information request to the Research Executive Agency, lists the many potential venues for “collaboration” through “synergies” with previous and completed EU-funded projects — as “clustering with other research projects in the field is considered “an essential and important tool for improving collaboration between researchers within Europe”.
Gained, exploitable knowledge and experience include:
a) “algorithms sharing, fusion and data analysis (where FOLDOUT, IPATCH, EFFECTOR, I2C and ANDROMEDA can share their lessons learned and know-how on information fusion and sensory data augmentation, on object detection and classification, and on high level modelling of behaviour analysis, with FOLDOUT’s ML predictors, when combined with all aforementioned competencies, can enhance EURMARS’s ability to effectively detect, classify and predict/prevent threats)”
b) “the shared intelligence, situational awareness and threat assessment (where IPATCH and FOLDOUT monitor around and outside the immediate border area pre-events and provide alerts using DSS tools, EFFECTOR and ANDROMEDA emphasize the use of platforms like CISE and EUROSUR and have the primary objective of monitoring and assessing real-time threats, and OCEAN2020 uses the “system-of-systems” approach to enhance situational awareness. All these projects include technologies that can be leveraged by EURMARS to enhance efficiency and effectiveness on all three of the aforementioned project’s aspects)”
c) “past trials, demonstrations and impact assessment (where OCEAN2020, EFFECTOR and ANDROMEDA can offer real-world pilot insights – including the leveraging information gained from research projects like PERSEUS, CloseEye, MARISA, and RANGER) as they emphasize in diverse maritime environments, and thus provide EURMARS with important lessons-learned regrading live and simulated trials as well as in translating results into scientific and commercial exploitation approaches”.
Other similar research projects “that could provide collaboration opportunities” are MEDEA, CALLISTO, AI-ARC, EFFECTOR and NESTOR. This means that “Since these projects are ongoing or running in their final stage, the EURMARS project will exploit their outcomes and current research to foster collaboration and leverage its own research results, exchanging methods and information”.
D6.1 also notes that “EURMARS partners are participating in several other projects, to further enhance interoperability and intelligence and, subsequently, to increase the quality of outcomes by collectively streamlining the factors that guarantee security for all EU citizens”.
Deliverable D1.3 (‘Ethics and Innovation Management’) contains some informative passages on the foreseen project pilots. This, for example:
“As EURMARS technology takes shape, it will be tested in pilot scenarios now being developed (Task 5.1 ‘Pilots Definition: Scenario, Methodology and Test Plan’). These pilot scenarios will deploy prototypes of EURMARS technology to test its data-collection capabilities and procedures, its performance and in fictional threat scenarios and gather user feedback (Task 5.2 ‘Living Lab in Bulgaria’, Task 5.3 ‘Demonstration in Cyprus’, Task 5.4 ‘Demonstration in the UK’, Task 5.5 ‘Cross-border Demonstration in Bulgaria/Romania’, Task 5.6 ‘Evaluation, Benchmarking & Lessons Learned’). The ethics risks related to this pilot research are minor, since the human participants will be employees of EURMARS partners. Nevertheless, the Ethics and Innovation Management plan in this document includes safeguards for protecting human participants, safeguarding human rights, and processing Personal Data.”
No deliverables, publications, or presentations on the website at the time of writing, while only 3 public deliverables have been published on Cordis so far — there were none however before we asked the Research Executive Agency why some deliverables should have been public already according to the project’s own timeline still weren’t months later, and got them released.
In other words, 18 months into the project, 4 deliverables that should have been public — and “automatically” so according to the redacted Grant Agreement we managed to obtain — for at least 6 months, and for 15 months in one case, were actually still not shared with the public until our request.
In June 2023 EURMARS held ‘The Ethics of Border Management Surveillance Technology’, an online workshop led by Trilateral Research to discuss the “ethical, social, and legal implications of an EU Border Management Surveillance System”. Topics discussed include:
“- EU Border Management: challenges and path going forward,
– Overview of the EURMARS Project and its objectives,
– Ethical and Social Implications of EU border Management Surveillance Systems,
– Legal Framework and Best Practices for Border Management Surveillance Systems,
– AI Act and its impact on border management surveillance technologies,
– Best Practices for Border Management Surveillance”.
AlgorithmWatch participated — and noted that, no matter how many times the organizers tried to gather inputs and feedbacks on positive use cases and best practices observed for border surveillance systems, none of the participants could come up with a single example.
One of the organisers tried to share an example of good practice — but of better international cooperation, not of a surveillance system — from Romania, but then AW asked them to instead share a specific example of a good comprehensive system for border management and surveillance, or a best practice of testing/deployment — but not even they could find a single one.
Actually, one of them openly admitted to be sharing privacy and human rights concerns, and enjoyed the idea of discussing “co-design practices” to be implemented in the future. Notwithstanding, one of the organizers claimed that EURMARS aims at becoming a role-model for further border surveillance projects.
Deliverable D1.3, in fact, claims that “ the EURMARS consortium took on the responsibility to ensure that the project is compliant with high standards for research ethics (including the Horizon Europe Ethics Appraisal Procedure); data protection; human rights; and the assessment, minimisation, and mitigation of other risks”, even going as far as providing a “human rights impact assessment for EURMARS technologies” in the forthcoming deliverable D7.4.
Lastly, EURMARS adopted the approach on ethical AI developed by the EU’s High-Level Expert Group on Artificial Intelligence, and — according to D6.1 — the Consortium will even “conduct foresight compliance work anticipating the entry into force of the proposed AI Act”.
As an important final methodological note, all the EURMARS deliverables we obtained — namely, its Grant Agreement and deliverables D1.3 and D6.1 — were still “subject to review and approval of REA”, wrote the EU Commission agency in response to our access requests, and should therefore “be considered as drafts and the final version might differ, in some cases even significantly.”
