ROBORDER aims to deal with the complexities of patrolling and protecting EU borders through an “intelligent holistic,” “unique” technological “solution,” which is ambitiously described as “a fully functional autonomous border surveillance system.”
This will feature “unmanned mobile robots, including aerial, water surface, underwater, and ground vehicles that will operate both independently and in swarms, incorporating additional sensors as part of an interoperable network.”
The aim is exceedingly solutionist: “To provide a complete and detailed situational awareness picture that supports highly efficient operations,” in a context in which “the heterogeneity of threats, the wideness of the surveyed area, the adverse weather conditions, (…) the wide range of terrains” and, most of all, the lack of innovative technological and “intelligent” solutions would make it otherwise impossible.
On a general level, the “main objective of ROBORDER is to develop a fully-functional autonomous border surveillance system with unmanned mobile robots including aerial, water surface, underwater and ground vehicles (UAV1, USV, UUV and UGV), capable of functioning both as standalone and in swarms, and incorporate multimodal sensors as part of an interoperable network.”
Various sensors will be comprised in a network that includes “enhanced static networked sensors such as border surveillance radars, as well as mobile sensors customised and installed on the vehicles” — among them, “(i) passive radars that can extend the capabilities of the existing border surveillance radars, (ii) passive RF (radio frequency, ed.)-signal sensing devices to intercept emission sources that are present in area, enrich the overall situational awareness picture with this information, allowing for further characterizing the nature and behaviour of entities in the picture, and detecting unauthorized signal sources and (iii) other mobile sensors like thermal cameras (infra-red), optical cameras and more.” (Ed.: Even the project’s own description have difficulty keeping track of all of the surveillance devices to be included.)
“Supplementary technologies” will also be developed to allow “robust” communication between the robots and a “command and control unit,” which will parse through “large volumes of heterogeneous sensor data” to promptly provide “a quick overview of the situation at a glance to the operators, supporting them in their decisions” — involving, for example, the “early identification of criminal activities and hazardous incidents.”
Automation is obviously key to the ROBORDER solution. The Cordis page says for example: “Additional command and control functionalities will allow for the translation of the intention of the operators into remote actions, automatic selection of the most appropriate composition of hardware (robots, sensors and communication links) for each given situation and easy deployment and operation of the fleets of heterogeneous robots and sensors.”
ROBORDER and other EU-funded research projects in border management and surveillance are deeply tied to each other.
Deliverable D7.7 (Report on standards and collaborations) for example says: “To facilitate the transfer of knowledge on key ROBORDER areas, especially surveillance, monitoring and security in multi-domain environments and the adoption of innovative technologies and approaches, ROBORDER partners collaborated with 20 other projects, including:
• 15 European Horizon 2020 projects;
• 2 European Defence Agency projects;
• 1 European Space Agency project;
• 1 European structural and Investment funds project;
• 1 Federal research project funded by the Deutsches Zentrum für Luft- und Raumfahrt German Aerospace Center.”
Non-EU countries were involved in these collaborations, namely Israel and South Korea.
The collaborations “covered the four domains of interest for border surveillance, i.e., sea, land, air, and underwater in the area of surveillance (52,9% of application areas), security and safety (31,4%), monitoring and management (10%), and autonomy (5,7%).”
Thematically, they “targeted situational awareness (34,8%), border management, security and surveillance (10,1%), integrated command, control and coordination platforms (15,9%), Law Enforcement Agency operations (7,2%), maritime surveillance (10,1%), airspace security and safety (7,2%), port of the future and port surveillance (7,2%), electronic warfare (2,9%), land use and urban environment (1,4%), circular economy (1,4%) and the topic “drive to the future” (1,4%).”
A complete list of projects with which ROBORDER collaborated is included in FIG.13 of the deliverable. It includes the H2020 projects ARESIBO, ANDROMEDA, AUTOPILOT, CAMELOT, COMPASS 2020, CONNEXIONs, COREALIS, CREST (focussed on crime/terrorism), RANGER, RAWFIE, ILEAnet, GAUSS, ALADDIN, iNGENIOUS, and iPROCUREnet.
A very useful list of “topics” concerning all such projects is also provided. Objectives and use cases for projects related to ROBORDER are included in Table 6.
Deliverable D8.5 further adds: “The platform of ROBORDER is based to already existing swarm operation systems that have been tested successfully in previous projects such as NOPTILUS and RAWFIE projects.”
ROBORDER is part of the BES Cluster of EU-funded projects in Border External Security, lead by METICOS.
The project’s results were further used in the REACTION project. Funded through the Border Management and VISA Instrument (2021-2027), and coordinated by the Greek Ministry of Immigration and Asylum, REACTION promises yet another “fully functional, next-generation, holistic border surveillance platform implemented through the use of unmanned vehicles,” including in swarms and with the ability to “continuously improve.” The REACTION platform will also be capable of providing “situational awareness from remote areas as an effective means of early detection of critical situations.” According to Greek authorities, the system will be “integrated and interconnected with the already installed information platforms at the Reception & Identification Centers (KYT) and Evros Police Station,” and “in the long term, it is expected to expand the use of these technologies to all unmanned aerial vehicles of the Security Forces and the Armed Forces.” The Greek research center CERTH, which coordinated ROBORDER and is now working on REACTION, was asked by ZDF Magazin Royale and AlgorithmWatch which specific results or components of ROBORDER are being used in REACTION but failed to provide an answer.
Three main pilot use cases are briefly described on the ROBORDER website. We learn that the “developed border surveillance technologies and ROBORDER platform will be tested in three main pilot use cases, consisting of several scenarios covering multiple situations of violations. In each evaluation cycle, three different use cases will be tested in different terrains, in order to exhibit the wide applicability of ROBORDER over different scenarios.”
The first pilot use case concerns the “early identification and tracking of illegal activities.” “To provide a complete and detailed situational awareness in border territories, multiple scenarios will be executed,” which includes In detail:
“(i) Detecting unauthorized sea border crossing: The use case involves the monitoring of sea passages and islets in the Greek archipelago. Accumulated data from supervision devices will be processed along with information coming from sensors like passive radars and optical cameras mounted on a swarm of robots.
(ii) Detecting unauthorised land border crossing and signals from trespassers: The system will be deployed at the Bulgarian-Turkish borders in order to track and monitor illegal activities. Specialized sensors will be able to detect the emissions coming from microwave devices (cell phones etc.) and so possible human activities could be recognized.
(iii) Detecting unauthorised land border crossing: The developed autonomous systems will allow to patrol hardly accessible areas (tested on Hungarian territories) to optimize surveillance and control situation. Multiple UAVs and UGVs will be exploited for collecting relevant information in an early phase and properly inform the operational personnel.
(iv) Tracking high-tech smugglers: The ROBORDER technologies will be used to track and recognize smuggling activities in order to prevent the transportation of illicit products across the Estonian-Russian borders. (Ed.: Was Russia involved in any capacity?)
(v) Detecting the terrorist attack coming through cross border: Based on intelligence information about a possible action of a hostile organization, the system will be deployed on Romanian coastal areas and tested in detecting possible explosive devices.
(vi) Early and effective identification of passive boats moving offshore: The Italian port authorities will deploy the system to investigate a possible intruder from closer position to provide a more accurate situation assessment.
(vii) Tracking organized crime activity in remote border areas: Law enforcement authorities from Northern Ireland and Republic of Ireland will deploy ROBORDER technologies in order to undertake improved surveillance operations for criminal activities recognition.”
The second pilot use case will concern the “early identification and tracking of illegal communications.” A third one will focus on the “detection of pollution and other accidents occurred in the borders” (e.g. by detecting and tracking pollutants spilled at sea).
Some basic details about the actual tests are included in D8.5: “The scenarios related to the land and marine border threats have been operationally tested and evaluated in Portugal, Bulgaria, Italy and Greece according to the evaluation plan and associated PUCs. Based on the unauthorized border crossing and pollution detection scenarios an analysis and report on the experiences and assessment results from the use of the ROBORDER platform have been made.”
According to the deliverable, the “task is successfully concluded, meeting all expectations according to GA (Grant Agreement, ed.) without any deviations.”
The scenarios included: “Detection of multiple groups of trespassers passing the boarder on foot (a large group of 10 trespasser splitting in two groups of 5 people); Detection of a van/truck crossing the border; Detection of an UAV crossing the border; Detection of a GPS spoofing and communication jamming attack on border security UxV.; Detection of small vessel approaching the shore; Detection of a small vessel entering a port.”
“Full details of demos/operational tests” were provided in Section 2.1 of deliverable D8.6, which is however not available to the public and was partially obtained by AlgorithmWatch only through an access to information request. The section is however entirely redacted in the file we received (pp. 10-17), including in the Table of Contents. More generally, all references to “operational tests and demonstrations” are blacked out throughout the document.
D8.6 on the other hand does say that “humans are going to be involved either as interviewees or as participants. In both cases, two types of data are going to be collected: personal data and evaluation data.” It also adds that in general, “the Consortium will rely on people already employed by them and the use of their own contacts.”
A worrying remark: “ROBORDER will collect imagery and videos of participants as well as location data, which can be linked to surveillance and tracking, in order to evolve the detection capabilities of the system.”
Even though the project is extremely sensitive in terms of privacy and human rights, the project structure on the project’s website was more than 4 years after completion still missing the ninth Working Package on “Ethics”, which, as a “new ethics-related plan,” had been “submitted by CERTH to update the ethics requirements.” In fact, “matters such as recruitment, collection of informed consents, data protection and privacy were refined to better reflect the needs of the consortium ensuring, though, compliance with EU and national legislation.”
We tried to access the ethics-related deliverables, but could — very partially, through access to information requests — only access two: deliverables D8.6 (Ethical Code and Updates on Data Protection) and 9.5 (DU – Requirement No. 15, on dual use issues).
We could not understand much of those heavily redacted documents. D9.5 was redacted right from the Executive Summary, half of which was blacked out in the first disclosure. What was clear enough: ROBORDER is portrayed as “a project focused on surveillance, which aims at developing and demonstrating a border surveillance system,” and may therefore “raise concerns regarding a potential harmful impact on human rights.”
This is especially true given the ambitious (solutionist) goals set by the project, to be realized through its technological outputs: data gathered by its heterogeneous unmanned vehicles could “prevent an environmental crisis taking place at the borders,” for example.
D8.6 claims that “ROBORDER will be able” both “to prevent illicit activities taking place at the borders” and “to protect EU citizens from criminals that are acting either at their border or that are planning to act within them.” Ultimately, “ROBORDER acts at the border areas of the Union (which) gives it the ability to monitor larger scale upcoming attacks (e.g. terrorist attacks or possible affiliations with terrorist organisations) that could target a whole country.”
The many redactions operated by the EU Commission before sharing ROBORDER deliverables with us also made it impossible to fully discern what happened with the ethical assessment.
D8.6 explicitly says: “When drafting this deliverable, an additional report was requested from the EEA (External Ethics Advisor, ed.) in order to ensure ROBORDER’s current compliance with the European Regulations.” Even though the EEA “considers the ethical issues handled properly in this project,” “time given and resources available allowed only a limited review and assessment of the project’s ‘ethical’ status, including D8.6 and its Annexes.”
We understand that internal assessments were conducted mostly by police forces or police-related entities — including from an illiberal democracy such as Hungary.
Annex III of D8.6 consists of a new deliverable, the EEA Ethics Report (D9.6, GEN – REQUIREMENT No. 19), which is labelled “confidential.” It interestingly contains a “Review of Deliverables,” which we managed to obtain in the clear after contesting initial redactions by the REA on public interest grounds. The EEA leaves some interesting critical remarks: “The information I have on the planned “Simulation Mock-up” is very limited” (D9.1); “too much text from regulations. More focus should be given on what it means for the project” (D9.2); the “solution” should not only “safeguard privacy” but also other ethical, legal societal, and political implications such a “solution” will have an impact on” (also D9.2).
At times, the opacity seems to be too much even for the EEA. For example, in D9.3 the EEA notes: “The existence of automated decision-making, including profiling (…) definitely needs specification and transparency to the data owner and the operator.”
The EEA notes in D9.4 that “requirements should not be limited to UAVs but also cover: autonomous border surveillance system with unmanned mobile robots including radars, aerial, water surface, underwater and ground vehicles… heterogeneous robots …command and control unit, decision support tools, etc.”
Importantly, a plethora of ethical risks — including to physical and mental health, on “societal cohesion and stability”, on “political conflicts” and on the environment — was identified for the ROBORDER project, both for the individual and society. The EEA stresses the urgency to address all of them, not just those related to data protection.
Dual use is also a contentious issue, according to the EEA: Terms such as “performance,” “misuse,” “adequate measures…” were rather vague and should be more clearly defined in ROBORDER. Also: “It is unclear why and how National Data Protection Authorities will be involved.”
The answer to the question “Were ethics issues correctly addressed by the project participants?” is consequentially “Partially”: “The Consortium has partly addressed the previous ethics requirements providing, inter alia, details of measures to address risks of dual use and misuse of the project’s outcomes. However, several ethics issues still remain.”
Among them: Although the Consortium has stated that no personal data will be collected, the informed consent form templates that have been provided include information regarding participants’ consent on the use of their videos and images. Several partners declare that they will collect personal data (D9.1 & D9.3). Detailed information on the involvement of personal data in the project needs to be provided.
“Other outstanding ethics issues” include:
a) an “explanation as to how the Independent this Advisor’s suggestions for addressing the ethics issues that arise from the project have been/will be taken into consideration;”
b) a “report from the internal Ethics Advisory Board, regarding how the ethics issues are dealt with in the project, as well as how the Ethics Advisory Board collaborates with the Independent Ethics Advisor;” and
c) an “evaluation of the ethics implications of the surveillance/profiling technologies produced by this project (including but not limited to misuse). The Independent Ethics Advisor has suggested performing an evaluation of the ethics implications of the project using an assessment methodology described as Ethical, Economic, Legal, Political and Societal Impact (EELPS) tool which the Consortium may take into account.”
Section 3.2.1 of D9.5, which we also managed to obtain in the clear only after contesting an initial non-disclosure, further adds: “The research and applied knowledge acquired in this project has the potential to be exploited by terrorists or criminal elements.” Given that “the project aims could be subject to dual use threat,” it is “of the utmost importance that a robust system is in place to ensure the work of the project is not exploited for subversive means.”
Overall, the “classification of ROBORDER results, the application of security best practices and the involvement of experienced security researchers (…) are considered sufficient measures to prevent the misuse of research findings.”
Solutionist jargon is rampant throughout project documents — starting from its overall promise: “Border authorities will have the capacity to handle large-scale crises in a timely manner, leading in this way towards an enhanced feeling of security among European citizens” (D8.5; this assumes that handling migration “crises” has to do with technological, rather than political solutions, ed.).
The project is clearly set on the needs of border agents, rather than of migrants: “The consortium has identified early border agencies’ needs thus, the assessment has been concluded with the required hardware components.”
At times, regulations seem to be considered as mere obstacles to invention. In deliverable D7.3, for example, a “market analysis” is performed, and one of the “three key elements for the drone market to further evolve” notes: “Regulations and social acceptance. There are lots of controversies about privacy and safety, which are strong constraints for applications already feasible from the technical point of view.”
Throughout the project, “ethics” mostly seems to concern how to make the developed technological solutions compatible with bureaucratic “ethics” requirements, but cannot — and structurally so — reflect on whether the idea of developing swarms of UAVs in the context of migration is ethical (fair, just, etc.) at all in the first place. This buys into the overall solutionist narrative according to which ethics and human rights are to be respected but first and foremost (“strong”) constraints to technological innovation.
D8.5 also contains an extremely concerning passage: “ROBORDER aims at providing a fully autonomous multi-robot systems, where human operators contribute only for inspection and supervision purposes. The heterogeneous robots are equipped with different sensing abilities that are coordinated by a plug-n-play framework and performs seamless transition between different surveillance tasks while operating. As a result, and by taking the existing static infrastructure into consideration, the operators are enabled to easily command and control the overall team of heterogeneous assets with the minimum workload” (p. 77).
What is left then of human autonomy and agency? Will a human still be meaningfully involved? And can a “fully autonomous” system such as the one imagined by ROBORDER be compliant with “EU values” at all? Too much information is still shielded from public view to provide evidence-based answers to such questions.
