Digital routes and borders in the Middle East: the geopolitical underpinnings of Internet connectivity

ABSTRACT

In the second half of the 2010s, the Arabian Peninsula experienced major foreign policy changes that suggested a reshuffling of the region as a security complex. This paper offers an analysis of the geography of Internet data routing in the Middle East in order to assess whether foreign policy shifts apply to cyberspace and the potential for meaningful cybersecurity cooperation. We specifically focus on the Qatar blockade and the Abraham agreements. Our main hypothesis is that the architecture of connectivity can reveal the level of existing cooperation between data-routing operators of different countries and can therefore serve as one among many indicators for the level of trust (or lack thereof) between these countries and, as a result, for the potential success of cyber cooperation. We argue that the architecture of connectivity only partially reflects foreign policy shifts in the Gulf Region and despite the political announcements, digital borders and rivalries in the region remain deep and could potentially create obstacles for meaningful cybersecurity cooperation.

KEYWORDS:

• Border Gateway Protocol
• Middle East
• digital borders
• cyberspace
• Gulf
• data routing
• Internet infrastructures

1. INTRODUCTION

In August 2020, Israel, the United Arab Emirates (UAE) and the United States signed a joint statement that marked the first public normalization of relations between Israel and a Gulf country. The Abraham Accords signed in September 2020 officialized the Israel–UAE agreement and the Israel–Bahrain agreement that were soon followed by the normalization of Israel’s relations with Sudan and Morocco. This historic breakthrough highlights the major foreign policy changes experienced by the Arabian Peninsula in the second half of the 2010s and a reshuffling of the region as a security complex. It raises numerous questions about the potential cooperation between these states in security and defence matters. Specifically, observers have identified the cybersecurity domain as one of the key items on this agenda, given the common concerns from Israel and the two Gulf States regarding Iran’s influence and hacktivism in the region (Jones & Guzansky, 2019; Rabi & Mueller, 2017; Ulrichsen, 2016).

Yet we know very little about how geopolitical alliances and rivalries, as well as foreign policy shifts do actually translate in the cyber domain. Do political alliances lead to cooperation in sensitive areas such as Internet connectivity? Do Middle Eastern countries share the same borders in cyberspace as they do in the physical world? This paper offers an analysis of the geography of Internet data routing in the Middle East in order to assess whether foreign policy shifts apply to cyberspace and the potential for meaningful cybersecurity cooperation. Our main hypothesis is that the architecture of connectivity can reveal the level of existing cooperation between data-routing operators of different countries and can therefore serve as one among many indicators for the level of trust (or lack thereof) between these countries and, as a result, for the potential success of cyber cooperation.

The research question we ask is the following: Does the architecture of connectivity reflect foreign policy shifts in the Middle East over the past decade? It raises a related methodological question: How can the analysis of the topology of the network inform fieldwork through the identification of relevant actors?

1.1. State of the art

This question sits at the intersection of several bodies of literature. First, it relates to the issue of measuring the gap between declarations of cooperation among states and the reality of that cooperation. International Relations scholarship has pointed out the discrepancies between the statements of decision-makers eager to boast about their cooperation and the limited results of the actual implementation of those promises – a phenomenon sometimes described as a case of ‘organized hypocrisy’ (Egnell, 2010; Krasner, 1999; Stein, 1990). The Arabian Peninsula is a case in point: cooperation among the member states of the Gulf Cooperation Council (GCC) in the defence and security domains has been extremely modest since the creation of the GCC in 1981 (Barnett & Gause, 1998; Sadiki & Saleh, 2020; Abdulla, 1999). Our methodology can help identify a disconnect between political speeches and the actual cooperation in data exchanges between the countries under study.

Second, it pertains to the science and technology studies literature, given the diversity of actors involved in routing policies and the complexity of their relationships. The in-depth analysis conducted by Oever (2020) of the sociology of socio-technical actors of Internet governance, and the processes by which they adopt norms to develop policies and standards, offers an important element to understand the co-construction of routing policies involving state and non-state actors. In particular, Musiani (2022) showed that the concept of digital sovereignty could be studied via the infrastructure-embedded ‘situated practices’ of different political and economic projects aiming to ‘establish autonomous digital infrastructures in a hyperconnected world’. The design of an architecture of routing that allows geopolitical control falls under this category. It is part of the ‘turn to infrastructure in Internet governance’ (Musiani et al., 2015). But control often comes with a cost for the resilience of networks. Indeed, simple and centralized networks are easier to control (Roberts et al., 2011) than complex distributed networks but they present many points of vulnerability to cyber and physical attacks and higher risks of congestion, making them less resilient. This trade-off between control versus resilience often reflects the strategic priorities of states and their perception of cyber threats.

In the Gulf, relations between public and private sectors tend to reflect old arrangements between the rulers and the merchant families (Crystal, 1990; Field, 1985; Herb, 1999; Murphy, 2006). Because kings and emirs needed the allegiance of the business class to build the newly independent states, the private sector in the region was largely constrained by the political priorities. Furthermore, the rentier-state model that prevailed in those countries, thanks to their oil and gas reserves, made the government the primary, if not the only actor of the economy, preventing the emergence of a truly private sector (Beblawi & Luciani, 2015). Our paper brings new methodologies to assess how and whether the alignment of state and non-state actors helps states achieve strategic and foreign policy objectives. It could also highlight how this alignment can constrain decisions by routing operators that can potentially run counter to their interests in terms of performance and/or resilience of the network.

Third, this paper brings a methodological contribution to open-source research, which is increasingly understood as a useful – if not necessary – part of academic research in social sciences. According to Limonier and Audinet (2022), open-source research should be considered a fieldwork in its own right as it allows for uncovering complex dynamics that are hard to identify through conventional sources. And digital fieldwork is particularly useful to prepare and organize a standard research fieldwork before it takes place. As such, the cartography of Internet routes based on open-source data can provide a clear understanding of the architecture of connectivity and its geopolitical underpinnings in a given region but also help identify key actors and stakeholders. Our study provides a panoramic view of the Gulf Region interconnectivity and its links to Israel, and offers unique insights into the key players in the region that allow a better-informed fieldwork on a sensitive security topic.

1.2. Argument

As the following discussion evidences, the architecture of connectivity only partially reflects foreign policy shifts in the Gulf Region. First, given the Abraham Accords, we could have expected stronger links between Israel and the Gulf countries. Our results show that Israel has now a direct connection with Bahrain but no connection with any of the other Gulf countries. There are also significant discrepancies between the narrative of an Emirati–Saudi rapprochement and the way both countries exchange digital data. We argue that despite the announced foreign policy shifts, digital borders and rivalries in the region remain deep and could potentially create obstacles for meaningful cybersecurity cooperation.

Likewise, the idea of Qatar isolated from its neighbours as the result of the 2017 blockade also appears exaggerated when looking at the interactions among the Internet networks of Gulf countries. The few direct links between Qatar and Saudi Arabia existing in 2015 were indeed cut-off in the midst of the blockade, and the overall number of routes from Qatar to Gulf countries sharply decreased, consistent with the foreign policy shift. But direct links persisted between Qatar and Bahrain, and between Qatar and the Emirates, yet through major government-controlled operators.

Third, from a methodological standpoint, our results highlight the central role acquired by Bahrain in the interconnectivity of Gulf countries and with Israel. Our methodology allowed us to uncover a major change in the architecture of connectivity of the region with the emergence of new actors, and thus provides crucial insights for fieldwork, a necessary additional step to better interpret our results.

1.3. Organization of the paper

In order to understand how foreign policy shifts might be reflected in cyberspace, we provide some geopolitical context in section 2 to explain the patterns of cooperation and competition in the region, with a focus on the Qatar blockade and the Abraham Accords. Section 3 presents the methodology used to assess the level of actual routing cooperation between Gulf countries and Israel. It describes how the Internet works at a logical level, through the interconnection of approximately 100,000 independent networks called autonomous systems (ASes) using the Border Gateway Protocol (BGP) to communicate. These interconnections create the mesh of routes available for digital data to transit. We further expand on how we collect data and detail our methodology that combines the visualization through graphs and the analysis of the architecture of connectivity in the Gulf Region and Israel, along with a fine-grained qualitative study of relevant features of connectivity. Sections 4 and 5 present our analyses and findings based on nine distinct graphs outlining different patterns of regional connectivity. Our work combines a diachronic and multi-scalar approach. We explore connectivity at the intra-regional and global levels, with a focus on Qatar, and its evolution between 2015 and 2021. Section 4 demonstrates that there are very few routes interconnecting Gulf countries between themselves and that these routes nearly all go through Bahrain. It also shows that the impact of the blockade has further reduced the number of intra-regional connections for Qatar and cut off its Internet routes to Saudi Arabia. Section 5 examines the evolution over time of the routes between the Gulf countries and Israel in the context of the Abraham Accords, comparing the architecture of connectivity in 2015 and 2021. It demonstrates that Bahrain has emerged only recently as a central node, and that the first interconnection between Gulf countries and Israel dates back to 2019. In section 6 we add to our cartography all the other countries that are directly connected to the Gulf countries and demonstrate that the Gulf intraregional connectivity is mostly carried out by major foreign actors. In conclusion, in section 7 we discuss a series of hypotheses to explain these results and the key importance of fieldwork facilitated by the identification of the most relevant actors. We argue that despite political declarations and commitments about cyber cooperation, the level of cooperation between data routing actors remains very low both between countries of the Gulf Region and with Israel, revealing the persistence of strong digital borders.

2. PATTERNS OF COOPERATION AND COMPETITION IN THE GULF REGION

2.1. Tensions in the Gulf, leading to the Qatar blockade

At the political level, the first major regional trend has been the tensions between Riyadh, Abu Dhabi, Manama and Doha that grew in earnest after the Arab uprisings of 2011. Whereas the Qatari leadership openly supported the new revolutionary leaders in countries such as Egypt, Libya or Tunisia, rulers from the three other Gulf States observed with deep suspicion the political transitions in the Middle East, fearing that this could foreshadow their own demise – a fear exacerbated in Bahrain as the ruling family confronted its own uprising during that same period (Kamrava, 2012; Ulrichsen, 2020). Soon, the public display of Qatar’s endorsement of Islamist movements affiliated with the Muslim Brotherhood, such as Tunisia’s Ennahda and the Freedom and Justice Party in Egypt, became a contentious issue within the GCC itself. Saudi and Emirati suspicions towards Qatar exacerbated as the latter also deepened its partnership with Turkey. Turkey’s Prime minister – and President after 2014 – Recep Tayyip Erdogan, also favoured Islamist forces in the midst of Arab uprisings. His political movement, the Justice and Development Party, was itself a Turkish offshoot of the Muslim Brotherhood and his decision in 2014 to open a military base in Qatar compounded the fears in Riyadh and Abu Dhabi.

Historically, both Saudi Arabia and the UAE had hosted activists of the Brotherhood as they fled Egypt in the 1960s. They even played a significant role in shaping local social life, from religious speeches to education programmes in schools (Freer, 2018). However, as the so-called Arab Spring emerged in 2011, ruling families in Riyadh and Abu Dhabi were intending on reducing that influence and did so through a wide array of coercive measures (closing of schools, banning of associations and political organizations, as well as deportation, and arrest). This culminated into two interrelated developments: the deterioration of their relations with Qatar and, conversely, the strengthening of bilateral ties between Saudi Arabia and the UAE.

Then on 5 June 2017, Saudi Arabia, the UAE and Bahrain suddenly announced the suspension of diplomatic relations with Qatar and the closure of their airspace and sea routes (Ulrichsen, 2020). This was justified by an intensive information campaign on medias and social networks, orchestrated primarily by the Emiratis, to portray Qatar as a state sponsoring terrorism and endangering the region (Jones, 2019).

Although tensions among Gulf States had occurred repeatedly since their independence in the 1970s, the intensity of the 2017 crisis was unprecedented and questioned the very existence of the GCC as a regional body gathering like-minded states (Sadiki & Saleh, 2020). This was compounded a few months later by the decision of Saudi Arabia and the UAE to formalize their bilateral alliance with its own mechanisms of consultation (Wintour, 2017). The fact that both countries announced the institutionalization of this new framework a day after a GCC Summit of Heads of States disastrously ended only after a few hours of talks underlined the idea that these developments were not mere diplomatic posturing but could change the way Gulf countries interacted and conducted their regional policies (Al Wasmi, 2017).

2.2. Normalization of relationships with Israel, leading to the Abraham Accords

In August 2020, the decision of the UAE– and a month later, of Bahrain – to establish diplomatic ties with Israel followed the same logic. Historically, Gulf States adopted the common Arab policy against Israel. Although Saudi Arabia provided limited military contribution to the 1967 war, it was after 1973 and the use of the oil embargo as a retaliatory measure for Western support to Israel, that Gulf States played a significant role in the conflict. During that period, Gulf kingdoms opened Israel boycott offices meant to prevent any exchange with the Jewish State. In the meantime, local media and academic textbooks often reflected anti-Semitic tropes.

However, that Gulf defiance towards Israel evolved around the 2000s. Gulf States had been in close contact with Israel for at least a decade. Qatar had actually opened an Israeli trade office in the 1990s, and discreet consultations regularly occurred between national security officials. These exchanges were neither secret nor publicly acknowledged, like other states in the past, Gulf States were benefiting from the ‘mistress’ approach in their relations with Israel: they were gaining from these security relations without suffering the cost of their publicization (Jones & Guzansky, 2019). In 2005, Bahrain was the first to close down its boycott office. The UAE also started toning down the anti-Israel rhetoric and in 2015 Abu Dhabi accepted the opening of an Israeli delegation to the International Renewable Energy Agency, based in the Emirati capital.

However, by 2020, the logic had changed and Emirati and Bahraini leaders assessed that they could afford the cost of public relations. Abu Dhabi did not seemingly consult its Gulf neighbours before its announcement. The documents constituting the Accords not only acknowledged the establishment of diplomatic relations between Israel and the UAE and Bahrain, it also detailed specific areas of cooperation such as finance, energy, defence industry and tourism.¹

The subsequent signature in September of that year of the Abraham Accords represented a departure from the traditional position of GCC states regarding the prospects of normalization with Israel. First, GCC states had traditionally followed Saudi Arabia’s leadership on the sensitive issue of public ties with Israel. At the Madrid Conference for Peace in 1991, the six members of the GCC were represented by the Secretary General of the organization, Abdullah Bishara, to signal their common view on the Israel–Palestine conflict. Later, they all endorsed the Saudi Peace Plan that became the Peace Initiative of the LAS in 2002. The initiative posited a collective recognition of Israel by Arab States in exchange for a full Israeli withdrawal from the 1967 occupied territories, the establishment of a Palestinian State with East Jerusalem as its capital, and the settlement of the refugee issue. Noticeably, none of these provisions were considered in the Abraham Accords that in practice involved no trade-off. In this context, the Emirati, and Bahraini decisions to sign these agreements indicated a desire of their leaders to break with the previous consensus and redefine their foreign policy framework.

Both developments – the Qatar crisis of 2017 and the 2020 Abraham Accords – were portrayed by pundits as ground-breaking. They were to be understood as illustrations of changing dynamics between the Middle East and the Gulf, specifically the decline of the GCC and the simultaneous rise of Israel–Gulf relations. Noticeably, cyberspace played a central part in both events.

2.3. Potential implications for cybersecurity cooperation in the region

These two foreign policy shifts have potential implications for cybersecurity cooperation in the Gulf Region. The crisis between Qatar and its neighbours soon took the form of a non-military conflict fought through the propaganda and disinformation apparatus of both sides. Social networks were the de facto battleground between Emiratis, Saudis and Qataris. Qatar’s websites were for the most part blocked from the Internet of its neighbours.

In the case of Israel’s relations with the UAE and Bahrain, it was widely assumed by commentators and researchers (Handler 2020) that cooperation in cybersecurity had been and would be a major component of these new relationships. Israeli companies in this domain are expected to play a major role in the Emirati and Bahraini cybersecurity markets. Only a few days after the signature of the Abraham Accords, Israeli and Emirati heads of cybersecurity convened a public (virtual) conference. Igal Unna from the Israeli National Cyber Directorate emphasized ‘we are threatened by the same threats … because of the nature of the region, because of the nature of our new, “outed” relations and because of who we are– strong economically and technologically’ (Reuters, 2020). Then, in April 2021, the UAE Signal Intelligence Agency acknowledged having cooperated with its Israeli counterpart in the context of a cyberespionage operation targeting Lebanese Hezbollah (Haaretz, 2021).

Altogether, these statements suggest that cyberspace has become over that period a true reflection of the new security arrangements in the Gulf and the broader Middle East. Along that logic, the observation of cyberspace should evidence the rapprochement between Emirati, Bahraini operators and their Israeli counterparts. It should also reflect the extent of the Saudi-led blockade through, for example, the disconnection of Qatar from the rest of the Arabian Peninsula. Such observations would then demonstrate that activities in cyberspace are the mere continuation of foreign policy.

Against that backdrop, our research tested the validity of this assumption: Did the cooperation between operators for data exchange through the Internet truly mirror, or at least capture, strategic trends within a region, that is, the Gulf and Israel? To answer this question, we used a methodology developed by the GEODE² research centre, combining the elaboration of dynamic AS graphs with geopolitical analysis to provide an understanding of the dynamics of Middle East interconnectivity.

3. METHODOLOGY

3.1. What is BGP?

The Internet is a network of networks of different size and importance, owned by public or private institutions (university networks, business organizations, Internet service providers (ISP), etc.). These entities are called autonomous systems (ASes) because they choose their own routing policies: each AS has full control and authority over the internal routing within its network and over the access policies for traffic transiting through its network. The interconnection of ASes creates the paths that are available for digital data to transit, thus forming a worldwide mesh of paths that constitute the Internet. As of November 2021, there were around 122,000 ASes (Potaroo, 2021), each identified by a unique number. Each AS owns and manages a set of contiguous IP addresses (or prefixes) allocated by a regional internet registry (RIR).³

These entities must cooperate to deliver packets from their source to their destination. Indeed, in order to travel from one part of the globe to the other, data packets usually have to transit through several independent ASes. AS administrators therefore have to decide which ASes to cooperate with and which routes to forward their data across the Internet. They establish two main types of agreements: customer-to-provider, and peering. Customer-to-provider agreements are mostly passed through ASes of different sizes. For example, an ISP of national importance will agree to pay a transnational or international transit provider that owns cables to access the rest of the Internet. Peering agreements are generally passed by two ASes of comparable importance, and imply that the two contractants share approximately the same amount of data between each other. These relationships constitute the routes between ASes available for data to travel across the network of networks. In most cases, these agreements are confidential, but our methodology allows us to infer these links thanks to the data we are able to collect.

There are different types of ASes, they differ in size and role in the global network. The interconnection of ASes produces a network that is both decentralized and pyramidal. Figure 1 offers a schematic view of this architecture. Here we use a taxonomy of ASes developed by the Center for Applied Internet Data Analysis (Luckie et al., 2013). At the top, Tier 1 ASes are transit providers. Transit providers constitute the core of the network (or the backbone). They are usually worldwide operators who own submarine cables and are highly interconnected between themselves through peering relationships. Tier 2 ASes are usually smaller transit providers or ISPs of various sizes. They buy access to the global Internet from transit providers and make their customers pay for this access. Finally, Tier 3 ASes, also called ‘stub ASes’, are at the edge of the network; they mostly host data and contents. Examples of Tier 3 ASes include cloud providers, content delivery networks (such as Akamai), but also private companies (e.g., banks, manufacturers) and public institutions (e.g., universities). In Figure 1, content hosted by a content delivery network (at the bottom left) will have to transit through several ASes that have either customer or peering relationships to reach the computer or smartphone of a user (bottom right). While there are many exceptions to this simplified representation, this example helps understand the fundamental architecture of ASes relationships and data paths.

Figure 1. Schematic illustration of the basic structure of autonomous systems (ASeS).

A schematic graph showing different types of autonomous systems (ASes) (ISPs, transit providers, content delivery networks, enterprise network), and their possible interconnections (peering or customer-to-provider). The ASes are structured as a hierarchical pyramid where Tier 3 ASes are at the bottom, Tier 2 in the middle and Tier 1 at the top.

3.2. Data collection and analysis

In order to exchange traffic, an AS has to interact with its neighbours and therefore needs to communicate with them. To do so, ASes use the BGP. They exchange two main types of announcements: update messages that detail an AS path, that is, the list of ASes to cross in order to reach a destination (IP address); withdrawal messages, signalling that an IP address is no longer reachable through the announcing AS.

We are able to capture these messages that allow us to deduce the relationships between ASes. For example, when an AS announces a path that crosses consecutively AS1 and AS2, we can infer that there is a link between these two ASes. We represent our results through a graph showing the links (BGP agreements) between the nodes (ASes).

To collect the data, we use a BGP observatory, a platform developed by GEODE that captures and processes the path updates advertised by the routers running BGP to update neighbouring routing tables (Roughan et al., 2011). The platform generates every minute a full graph of ASes relationships, obtained by processing up to 30 BGP flows from publicly available BGP routing data, RouteViews,⁴ and the RIPE Routing Information Service (RIS),⁵ which aggregates BGP messages from BGP monitors at cooperating ASes (Orsini et al., 2016). This real-time snapshot contains about 89,000 nodes and 200,000 links. In addition, we augment BGP announcement data by adding relevant information such as the name associated with each AS, the country where the AS is registered, the number of IP address prefixes announced by the AS and the number of times a connection has appeared on the routing table, that is, how many times an AS has appeared in the collected BGP advertisements. Finally, we use the Potaroo blog to obtain statistics about the number of prefixes and ASes associated with each country year after year (Potaroo, 2020).

3.3. Limits

Empirical studies based on BGP data face a number of methodological challenges caused by the highly dynamic nature of the Internet and the high frequency of technical incidents affecting routers which can fail and be restarted at any given time. Information furthermore changes frequently as ownership of ASes and relationships between ASes evolve quickly. Routers thus have to deal with a continuous flow of update and withdrawal messages (up to 12,000 BGP announcements per second). Our BGP observatory takes this frequency into account by collecting routes over a long period of time (ranging from several days to several years), thus allowing us to discard short-lived changes that are not structurally significant.

These AS graphs are however known to be incomplete. Most notably, BGP path-filtering policies do not expose less-preferred paths that would be chosen if the preferred announced paths were not available (Gregori et al., 2012). Our collection methods, however, provide among the most comprehensive datasets available in open-source as of 2021.

3.4. The geopolitical significance of BGP

BGP is geopolitically significant in many ways (Douzet et al., 2020). These agreements are guided by technical criteria – usually the shortest route in terms of number of crossed AS (Chiu et al., 2015) – and economic choices (the cheapest route) but also by security and geopolitical concerns because the system is easy to manipulate for malicious or strategic purpose (Benton & Camp, 2016). In other words, the administrator of an AS determines its routing policies according to economic, technical and/or political considerations. Similarly, the decision of an operator to advertise a route that lets traffic cross its AS depends on its commercial policy, strategy and competitive environment, along with technical considerations. Thus, the paths available change according to trade agreements and competition between economic and/or political actors. Routing is therefore a field of friction between the different actors of the network. Graphs also help us understand the level of effective cooperation between the ASes of different countries in a regional context. Ultimately, this cooperation defines the routes and therefore the shape of cyberspace. In addition, the structure of connectivity can also be critical to the resilience of a network and create dependency relationships between territories, providing some countries or private actors with a form of influence or even topological power (Allen, 2011) over other territories.

3.5. Representing, observing and analysing data from BGP feeds

In this article we designed graphs to represent the architecture of connectivity. We followed several steps.

First, we collect full graphs of all the paths advertised, through our BGP observatory, as explained above, which provides us with a set of snapshots of all Internet routes.

Second, we apply filters to these snapshots in order to select the ASes relevant to our case study. For this article, we focus on the ASes registered in the Gulf countries and, for some of the graphs, Israel.⁶ We show two different types of graphs: intra-regional connectivity (between Gulf countries and with Israel), and global regional connectivity (between Gulf countries and all other countries they are connected to). We also provide a specific focus on Qatar in order to track connectivity changes in the context of the blockade. While Egypt was an important actor of the Qatar blockade, we have not included it in our study for two main reasons. First, this article focuses on the Arabian Peninsula as a coherent subregion, of which Egypt is not part. Second and more importantly, Egypt had no connectivity links with Gulf countries before 2014 and the deterioration of Egypt–Qatar relationships, which prevents us from establishing meaningful chronological correlations.

The graphs displaying intra-regional connectivity represent ASes registered in the Gulf Region (Figure 2), or ASes registered in the Gulf Region and Israel (Figures 6 and 7), as well as the links between them. This allows us to observe how these countries are interconnected and to reveal the structure of connectivity of the region along with the most central actors.

Figure 2. Internal connectivity of the Gulf Region, 2021.

The graph is made up of two big clusters. On the right, the autonomous systems (ASes) of Iran are separated from the rest; on the left, the cluster is made of five subclusters of the ASes of Iraq, Kuwait, United Arab Emirates (UAE), Saudi Arabia, and Bahrain–Oman–Qatar. They have few connections between them. At the centre, one big node from Bahrain is connected to numerous other ASes.

The graphs displaying global regional connectivity reveal how countries in the Gulf Region – and Israel– connect to the global Internet. In these graphs, we select not only the ASes from the region, but also their direct neighbours, that is, the other ASes registered anywhere else in the world with which they have a direct link through a BGP agreement. These more complex graphs include a high number of nodes, so we apply a filter that eliminates ASes with fewer than two neighbours. Indeed, these ASes are on the edge of the regional network and, by definition, play no role in interconnecting the region.

Third, we obtain graphs where each node is an AS, and each link is a BGP agreement or, in other words, a path available for digital data. We then colour each node according to the country of registration of the AS. In order to better visualize the level of interconnection between nodes, we use a visualization algorithm named Force Atlas 2 (Jacomy et al., 2014). This algorithm is commonly used in graph spatialization. It simulates a mechanical system of repulsion and attraction that makes nodes with a high number of connections closer to each other, while nodes with few or no interconnections are driven apart from each other. The application of Force Atlas algorithm gives us an image that transcribes the connectivity situation of the region, and that shows the interconnections between countries, or lack thereof.

Furthermore, we modify some of the visualization properties to make the graphs more easily readable. First, we force very concentrated hubs of nodes to be a little more scattered. We also prevent the overlapping of nodes, to be able to observe all of them. We then apply a different size to each node, using a metric of betweenness centrality (Freeman, 1979). The betweenness centrality of a node measures the proportion of the shortest paths between all nodes of the graph that go through this node. In other words, an AS located on many short data paths will have a bigger size.

With this method, we have created a set of different graphs that we present and analyse in the next two sections, in order to assess whether and how the Abraham Accords and the Qatar crisis are reflected in the architecture of connectivity of the Gulf Region.

4. TENSIONS BETWEEN GULF COUNTRIES AND THE QATAR CRISIS

The mounting tensions between Gulf countries over the past decade resulting in the Qatar crisis led us to formulate the hypothesis that the Saudi-led blockade could have cut off Qatar from the rest of the Arabian peninsula. In order to test this hypothesis, we studied the links between Gulf countries in order to understand the dynamics of intra-regional connectivity in 2021; then we focused on the interconnection of Qatar with the rest of the peninsula and its evolution in the context of the 2017–20 crisis.

4.1. Intra-regional connectivity in the Gulf Region

Figure 2 shows the data routes between the Persian Gulf countries as of 7 July 2021. In order to build this graph, we selected the ASes registered in eight countries: Iran, Iraq, Kuwait, Bahrain, Saudi Arabia, UAE, Oman and Qatar. We then looked at how they were interconnected with each other.

First, Bahrain is by far the most important regional stakeholder in terms of connectivity. Nearly all relationships between Gulf countries transit through Bahrain, and more specifically through one specific AS: AS 59605 (Zain). This node appears as the largest in the graph as it is the only AS registered in the region that connects to all the other countries. Indeed, nearly all data routes between the Gulf countries of our graph go through Bahrain.

Second, there are not many direct links between the Gulf countries. Outside AS 59605 (Zain), a few other ASes in our graph are connected to several countries in the region: the Emirati AS 8966 (Etisalat), the Kuwaiti AS 9155 (Qnet), the Iraqi AS 208293 (Moc-Alsalam), etc. However, very few routes converge through them and given their small size, they are most likely not significant in terms of Internet traffic.

Iran is connected to the region solely through AS 49666 (Telecommunication Infrastructure Company– TIC), a gatekeeper of Iran’s domestic network (Salamatian et al., 2021). This AS is connected to only two other ASes: AS 59605 (Zain) in Bahrain and the Emirati AS 49832 (Qbicomm). This Emirati AS is strongly embedded into the Iraqi network, and it serves mainly to connect Iraq to both Iran (AS 49 866) and Bahrain (AS 59605).

Iraq’s network is clearly centred around AS 208293 (Moc-Alsalam), which is directly connected to AS Emirati AS 49832 (Qbicomm) and, of course, to Bahrain through AS 59605 (Zain).

The network of Saudi Arabia is mostly connected to the region through three ASes, by order of importance: AS 39386 (STC-IGW), AS 35753 (ITC) and AS 35819 (Mobily). They are all connected to Bahrain’s AS59605 (Zain). However, in addition to Iraq, AS 39386 (STC-IGW) also connects Saudi Arabia to a few other small ASes in Kuwait and Bahrain, as well as to the Emirati AS 8966 (Etisalat).

The network of Kuwait is less centralized than most networks in the region, hence the lack of large Kuwaiti nodes. There are four main Kuwaiti nodes but only two are labelled in our graph: AS 21050 (Fast-Telco) and AS 9155 (Qnet). These four ASes are all connected exclusively to Bahrain’s AS 59605 (Zain), with the exception of AS9155 (Qnet), also connected directly to Iraq and Saudi Arabia through two ASes of each country, including two ASes embedded in Kuwait’s networks, which are too small to be named in the graph: Iraqi AS 202055 (QematAlwasat) and Saudi AS 42028 (EFG-Hermes-Kuwait).

Qatar and Oman share almost the same very centralized structure for their national networks. They are both connected to the rest of the region through one single major AS, which is connected to Bahrain’s AS 59605 (Zain): AS 50010 (Nawras) in Oman and AS 48278 (VodafoneQatar) in Qatar. Each country also has a direct connection to the UAE (AS 8966) through another AS: AS 8529 (OmanTel) in Oman and AS 8781 (QA-ISP) in Qatar.

Finally, the UAE shows a more distinct landscape, with a core network of ASes assembled as a cluster and connected to Bahrain– and a few other ASes– mostly through AS15802 (DU). But there are also a number of other ASes scattered across the graph. These scattered ASes appear to have two different types of functions: some of them are merely hosting companies and cloud computing services, situated at the periphery of the network, for example, the small brown nodes connected only to AS 59605 (Zain); the others serve as interconnection nodes, as is the case with the Emirati ASes in Iraq’s network. There are also a few other examples of UAE ASes fully embedded in other countries’ networks, such as in Iran (AS 40987-Vaulsys) or in Saudi Arabia, where AS 35086 (nourglobal) connects two major nodes of the Saudi network: AS 35753 (ITC) and AS 35819 (Mobily). However, they are not central in the connectivity, and their presence depends on reasons specific to each of them.⁷ One specific Emirati AS stands out: AS 8966 (Etisalat), which is connected to five different countries (Oman, Bahrain, Kuwait, Qatar and Saudi Arabia), making it the second most diversely interconnected AS in the region.

Figure 2 therefore clearly demonstrates that the level of interconnection between ASes of the Gulf is very low and for some countries totally non-existent. We therefore observe very low levels of regional cooperation. The few existing direct connections are not transit providers and they do not accept traffic from other AS beyond their end point, so they are not significant in terms of interconnection, and they serve specific purposes of their endpoints. These results overall reflect a strong desire to control data traffic by the countries in the region as there are very few pathways between Gulf countries.

The impact of the blockade has further reduced the number of regional interconnections for Qatar.

4.2. The evolution of Qatar’s intra-regional connections in the context of the blockade

We created three graphs that show the evolution of Qatar’s architecture of connectivity in the region. We took three snapshots that follow the chronology of the crisis between Qatar and its neighbours, choosing dates ante crisis (2015), during the crisis (2019) and post crisis (2021). Figures 3–5 display the ASes registered in Qatar and their immediate neighbours within the Gulf Region for these three different periods.

Figure 3. Connectivity of Qatar within the Gulf Region, 2015.

The graph shows the autonomous systems (ASes) of Qatar and the three among them that are connected to foreign ASes in the Gulf. Two of them are only connected to separate AS from the United Arab Emirates (UAE). The third one, AS8781, is connected to most of the other Qatari ASes, and to four ASes of Bahrain, three of Saudi Arabia and two from Kuwait.

Figure 4. Connectivity of Qatar within the Gulf Region, 2019.

This graph shows the autonomous systems (ASes) of Qatar and two among them have connections with foreign ASes in the Gulf. One Qatari AS is connected to one AS in the United Arab Emirates (UAE) and to one Bahraini AS, Zain. The other one is connected to one Emirati AS, one Bahraini AS and one Kuwaiti AS.

Figure 5. Connectivity of Qatar within the Gulf Region, 2021.

This graph shows the autonomous systems (ASes) of Qatar and the two among them with connections to foreign ASes in the Gulf. One is connected to one Bahraini AS, the other is connected to one Emirati AS, one Bahrain AS and one Kuwaiti AS.

The decrease in Qatar’s connections with its regional neighbours is visible in the graphs. Qatar’s gateways to Gulf countries are circled by a thick black line and the links to outside countries are also underlined by a thicker black line. Figure 3 shows Qatar’s links with Gulf countries in 2015. There are overall 10 data routes to four different Gulf countries. Qatar is indeed connected to two ASes in Saudi Arabia, in addition to ASes in Kuwait, Bahrain and the UAE.

The small cluster of nodes at the bottom left of the graph is disconnected from the rest of the network. Emirati network providers are linked to two ASes: Vodafone Qatar (AS 48728), the Qatari branch of Vodafone, a global mobile phone operator offering mobile service in Qatar; and to Gulf Bridge International (AS 200612), a regional cloud and connectivity provider. The purpose of Gulf Bridge International is to provide global connectivity to its particular clients whereas Vodafone Qatar provides connectivity within the Vodaphone network, that is, its purpose is not to convey the Internet traffic of the country. It is noteworthy that AS 200612 was registered in Qatar in 2015, in the UAE in 2019 and finally in the Virgin Islands in 2021, which explains why it does not appear in the 2019 and 2021 graphs. The main point of interconnection is AS 8781, Oreedoo Qatar (formerly Qatar Telecom).

In 2019, Figure 4 shows that the overall number of links to Gulf countries has dropped to five and the two links to Saudi Arabia have disappeared. Indeed, Qatar was connected to two major Saudi ASes in 2015 (35753– ITC and 41426 – STC) and has lost both of them in 2019. Qatar had not recovered any link to Saudi Arabia by 2021. Figure 5 shows that the number of links between Qatar and Gulf countries had further dropped to four in 2021. Qatar is directly connected to two ASes in Bahrain, including Zain (AS59605), one AS in the UAE (Etilsat, the national historical telecommunications operator) and one in Kuwait (Wataniya Telecom).

The Qatar crisis is therefore reflected in cyberspace by the drastic decrease of the number of data routes with other Gulf countries and the loss of its connections to Saudi Arabia. However, a limited number of links remain with Bahrain and the Emirates, despite the blockade, through major national operators.

5. THE ABRAHAM ACCORDS AND THE FEW DATA ROUTES BETWEEN THE GULF COUNTRIES AND ISRAEL

The context of normalization of relationships between Gulf countries and Israel led us to the hypothesis that the observation of cyberspace should evidence more cooperation between Emirati, Bahraini routing operators and their Israeli counterparts. In order to test this hypothesis, we added Israeli ASes to our graph of intra-regional connectivity in the Gulf Region to visualize how Israel was connected to Gulf countries in 2021. We then created the same graph for 2015 in order to assess the evolution over time and the potential impact of the foreign policy shift.

Figure 6 shows that all the connections between Israel and the Gulf countries transit through Bahrain. There are no direct connections between Israel and the other Gulf countries.

Figure 6. Internal connectivity of the Gulf Region and Israel, 2021.

This graph is similar to Figure 1, but with a third cluster for Israel. This cluster is on the opposite side of Iran, on the left of the Gulf cluster. Three autonomous systems (ASes) of Israel are connected to Gulf countries, only through the Bahraini AS of Zain.

Israel’s network is organized around three major ASes. Two of them are connected directly to AS 59605 (Zain) in Bahrain and concentrate most of the routes to Gulf countries: AS 1680 (Cellcom) and AS 8551 (Bezeq International). The third one, 12400 (Partner) is less central and also connected to Bahrain but through an Israeli intermediary, AS 9116 (Goldenline).

We hypothesized that this connection might be recent and linked to the new dynamics of cooperation in the region. To test this hypothesis, we looked at earlier data. Figure 7 shows the same graph realized in 2015.

Figure 7. Internal connectivity of the Gulf Region and Israel, 2015.

This graph shows three big clusters. In the centre there is a central cluster made of three subclusters: one of the Iraqi autonomous systems (ASes), one of the Emirati ASes, and one at the centre with ASes of Bahrain, Saudi Arabia, Qatar and Kuwait. On the right, a separate big cluster for Iran is connected to the centre through an Omani AS. On the upper-left side, very far from the other countries, the Israeli cluster is not connected to them.

The graph clearly shows that Israel is completely detached from the graph and shares no connection at all with the Gulf countries. By refining our analysis, we were able to date the materialization of a logical connection between Israel and Bahrain back to 2019. It also demonstrates considerable evolution over time in intra-regional connectivity. In 2015, Bahrain did not play such a central role in interconnecting Gulf countries. Furthermore, compared to 2021, the UAE’s network was less scattered in the region. At the time, the two most prominent internal hubs were Oman and Saudi Arabia. Oman’s AS 8529 was central in connecting Iran to the other Gulf countries. This AS belongs to OmanTel, the first telecom company in Oman. Saudi Arabia’s AS 41426 was central in the graph but has since been abandoned, that is, this AS is not active anymore in the network, and has probably fully ceased activity around 2018, the date of its last update on AS databases.⁸ As in 2021, Saudi Arabia’s AS 39386 had an important role, not only as a gateway between the region and the Saudi national network, but also as the only gateway in the region between Iraq and the other Gulf countries.

These graphs clearly show that there is no direct cooperation between most Gulf countries and Israel. The few routes connecting Israel to the Gulf countries go through a single AS located in Bahrain, AS 59605 (Zain). The Abraham agreements have not yet materialized in any substantial cooperation at the AS level, even between the Emirates and Israel.

The graphs also reveal that Bahrain has become a major point of interconnection within the Gulf and with Israel. The AS 59605 (Zain) has acquired a central position in the interconnectivity of the region and could therefore constitute a potential vulnerability. The central position of AS 59605 (Zain) means that a disruption to this AS would most likely shut down all the intra-regional data routes, leading the intraregional traffic to transit through foreign ASes outside the region. We therefore sought to assess to what extent foreign ASes outside the Gulf Region play a more central role in the interconnectivity of the region. In other words, if Bahrain’s AS 59605 (Zain) were to fail, what are the other alternative routes significant enough to avoid great disruption?

6. AN INTRA-REGIONAL CONNECTIVITY MOSTLY CARRIED OUT BY FOREIGN ACTORS

To test this hypothesis, we produced a set of graphs representing what we call the global connectivity of the Gulf Region as of 7 July 2021. The following graphs show the ASes registered in Gulf countries and Israel, and their direct digital neighbours, that is, the ASes from other countries to which they are connected. In other words, the graph helps us highlight the foreign ASes that serve as gateways to the rest of the Internet for the Gulf countries (and Israel) as well as points of interconnection between Gulf countries.

In order to make the graph easily readable, we removed the ASes that have fewer than two connections with other ASes. These ASes, located on the edge of the network, are generally Tier 3 ASes: private companies, cloud computing services, content delivery networks etc. They serve no specific purpose in terms of traffic transit and Internet access. The national networks of countries from the region are in cold colours (shades of blue, green, purple), and other countries with the most foreign ASes connected to the Gulf Region such as Germany (dark brown) or the United States (red) are in warm colours. For clarity, the other ASes are pictured in grey.

First, we notice that despite the addition of digital neighbours, AS 59605 (Zain) remains the biggest one in the graph. This means that most of the shortest routes transit through this AS. We also observe that this AS constitutes a major gateway between Gulf countries and foreign countries. We can infer that this AS not only plays a major role in interconnecting Gulf countries but also plays a notable role in connecting the region to the rest of the world.

Second, there are important actors other than Zain that play a role in interconnecting Gulf countries. These actors are for the most part Western third-party ASes. A considerable number of Western ASes are to be found at the centre of the graph. This position in the graph shows that they share a lot of links with other nodes. Their relatively small size indicates that they however do not concentrate on the shortest paths. These Western ASes actually are major transit providers with a worldwide reach. They are connected to most of the national gateways of each Gulf country, such as 49666 (Iran), AS 39386 (Saudi Arabia) AS 5966 (UAE), AS 9155 (Kuwait). All the national gateways of Gulf countries are linked together through at least two transit providers such as Level 3, Telia, Cogent, or others, which are among the largest⁹ providers in the world. This means that these countries mostly interconnect through third party actors, and specifically Western ASes.

We also notice in Figure 8 that Israel is nearly invisible, which can be explained by the only three links between Israel and Gulf countries shown in Figure 6. We therefore tried to assess more precisely how Israel is indirectly connected to Gulf countries. We therefore produced the same graph with Israel included. Figure 9 shows the graph with ASes from the Gulf Region and Israel, and their direct neighbours.

Figure 8. Connectivity of the Gulf Region and their neighbours, 2021.

The graph has at its centre a complex and diverse cluster of mostly foreign autonomous systems (ASes), with a few local important ASes from Bahrain, United Arab Emirates (UAE) and Saudi Arabia mostly. The cluster is surrounded with smaller clusters of ASes from different countries: Saudi Arabia, UAE, Qatar, Oman, Kuwait and Iraq. On the far right of the graph, most of Iran’s ASes form a separate cluster. One big AS from Iran, AS4966, is the only one very close to the Gulf and Foreign ASes.

Figure 9. Connectivity of the Gulf Region, Israel and their neighbours, 2021.

This graph is very similar to Figure 7. However, autonomous systems (ASes) from Israel form different subclusters that are scattered close on the upper-left part of the central cluster. Some Israeli ASes are mixed with the central cluster, and others are located between countries’ subclusters, such as between United Arab Emirates (UAE) and Saudi Arabia, or UAE and Kuwait.

The position of Israel is revealing in the way it does not change at all the structure of the graph. In terms of global connectivity, Israel is highly connected to Western and other foreign ASes that create pathways between Israel and the countries of the Gulf. In other words, Israel is mostly connected to Gulf countries through Western ASes intermediaries. As demonstrated in the previous section, Israel is directly connected to the region only through Bahrain’s AS 59605 (Zain). We have highlighted in black the only two connections that tie Israel to the Gulf countries, that is, between Israel’s AS 1680 (NV) and AS 8551 (Bazeq), and Zain.

In order to observe the evolution of the global connectivity of the Gulf Region, we created the 2015 version of the Figure 8 graph, showing the global connectivity of the Gulf countries. The situation in 2015 illustrated by Figure 10 is quite comparable to 2021, albeit several noticeable differences. First, the number of ASes is lower, which is consistent with the global trend of increase in the overall number of ASes since the 1990s.

Figure 10. Connectivity of the Gulf Region and their neighbours, 2015.

This graph shows a central cluster of mostly foreign autonomous systems (ASes) and big ASes from Saudi Arabia, Oman, United Arab Emirates (UAE), Iraq and Qatar. The clusters of the ASes of most Gulf countries surround it but they appear more scattered and mixed between them. On the right, the cluster of Iran is also somewhat scattered, but also closer to the central cluster, and it is connected to it through two different ASes.

Second, contrary to 2021, we can see a number of external– that is, outside the Gulf– ASes that have a significant importance in terms of betweenness centrality, mainly from Europe (in warm colours or in grey), such as 1299 (Telianet, registered in the European Union), AS 6762 (Seabone, in Italy). At the time, these nodes were central in interconnecting Gulf countries. In other words, the role of these few Western ASes was more important in 2015 than in 2021 in the interconnection of Gulf countries. The evolution between 2015 and 2021 thus shows two noteworthy dynamics. First, there is more diversity in terms Western ASes that provide external connectivity to the region. Western nodes are both smaller and more numerous in 2021 than in 2015. Second, we observe the restructuring of the intraregional architecture of connectivity towards the concentration of the shortest routes into a single central node.

Third, indeed, our 2015 graph reveals that there was no major central node for the whole region in 2015. Bahrain’s AS 59605 (Zain) did not play a role in the region’s connectivity, whereas in 2021 it has become central in the graph. However, in 2015 as in 2021, almost every national network has generally one or two gateway AS, that serve as access points for the country, and is connected to numerous ASes in the rest of the world, most notably to the United States and European countries. For Iraq, it is AS 44217 (IQnetworks), for Qatar AS 8781 (QA-ISP), for Saudi Arabia AS 39386 (STC-IGW), for Oman AS 8529 (OmanTel), etc.

Finally, we observe that Iran had a different single major AS (AS 12880, DCI) that was then the major gateway for most of the Iranian ASes towards the rest of the world. We also note that Iran’s network was closer to the global network in 2021 compared to 2015. This is the consequence of the evolution of the Iranian network. In 2021, the major gateway for the country had become AS 49666, a government controlled AS highly connected to international ASes but still poorly connected to Gulf countries through only two links to Bahrain and Iraq.

7. DISCUSSION AND CONCLUSIONS

Through our analysis of BGP data, we were able to unveil the architecture of connectivity in the region and shed light on a sensitive dimension of cooperation among Middle Eastern states. The trends identified only partially reflect the foreign policy shifts announced in the context of the Qatar Crisis and the Abraham Accords. Our study also leads to strategically significant findings that require fieldwork to be fully interpreted but can lead to some discussion.

First, the level of cooperation between data-routing operators is very low between countries of the Gulf Region. There are indeed very few direct connections between Gulf countries and most of them are not significant in terms of traffic transit, given the nature of the operators that provide the interconnection.

These results can first be interpreted as reflecting a strong desire by states to control data traffic in the region and a sign that they prioritize control over performance and resilience of the networks. Indeed, given the geographical proximity between the infrastructures and the level of interactions between countries for economic purposes, there is a good economic and technical incentive to provide multiple pathways between Gulf countries to avoid latencies and congestion. The absence of these links is significant. It reveals strategic choices made by operators, most likely guided by political distrust. Although economic routing operators and state actors are not necessarily aligned in their interests and decisions, they seem to be in the context of Gulf country connectivity.

Second, Internet routes between Gulf countries mostly go through either a single operator in Bahrain (Zain) or through foreign – and mostly Western– operators. Bahrain has emerged recently as a central point of interconnection in the region.

These results are difficult to interpret without field work. Open-source research and remote interviews tend to indicate that Bahrain has become a hub because Zain is run by a very dynamic operator belonging to the royal family of Kuwait with a proactive commercial policy to build agreements. But there could be other factors worth investigating, including the importance of the geographical location of Bahrain – an island between Qatar and Saudi Arabia– specific investment plans to build regional data centres¹⁰ or the role of international military agreements that can incentivize the emergence of a regional hub to facilitate traffic surveillance. Oman used to be the point of interconnection but could have lost its status when the new Europe–Persia Express Gateway (EPEG) cable was deployed in 2012, creating a direct link from Oman to Iran and Russia. This might have encouraged a progressive restructuring of BGP agreements in the region out of distrust.

The central position of AS Zain in Bahrain potentially creates a major point of vulnerability and also creates important strategic dependencies that would be interesting to document further with other sets of data, such as actual data flows or in-depth qualitative studies. Disconnecting the node that interconnects most of the ASes in the region could lead to disruptions. To test this hypothesis, however, we would ideally need to measure the volume of traffic that transits through Bahrain to assess its real importance in the interconnection of the region, which our methodology does not permit due to the lack of publicly available data. Indeed, there are also many pathways through international ASes

Third, most of the Gulf countries connect their network to the global Internet through a limited number of domestic points that are usually state controlled operators. This architecture of connectivity allows greater control of the routes data take but also creates political and technical vulnerabilities. Here the trade-off consists in connecting only to the operators that are indispensable to send their data across the global Internet, that is, major Western operators. Interconnection within the Gulf Region, however, is impeded by political tensions as illustrated by the Qatar crisis.

Fourth, in the context of the blockade, the number of regional Internet pathways for Qatar sharply dropped and the links to Saudi Arabia disappeared. However, direct interconnections with Bahrain and the UAE were maintained through a single node, under the control of the government, that is, under tight surveillance.

Finally, in the context of the Abraham Accords, cooperation with Israel is increasing but remains so far limited. Before 2019, there was no cooperation at all between the data-routing operators of Israel and the Gulf countries. In 2021, we observe a direct connection between three Israel ASes and a single operator in Bahrain, Zain, which has emerged as a central operator in the region.

These results only partially reflect the foreign policy shift linked to the normalization of the relations between Gulf countries and Israel and they show a clear discrepancy between political announcements and the reality of cooperation in the area of Internet routing.

Overall, we can argue that routing is a much more sensitive area of cooperation than others, including in the cybersecurity field where training or even equipment are more easily shared. Routing cooperation requires a very high level of trust as it is perceived as a domain of sovereignty and can also be manipulated for information – and population– control. For routing operators who seek the performance and resilience of their networks, the incentive is to have multiple interconnections. The fact that these interconnections do not exist reflects a political choice not to cooperate. This contributes to explain why the architecture of connectivity only partially reflects foreign policy shifts.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Additional information

Funding

This study was supported by the William and Flora Hewlett Foundation [grant number #2019-8617].

Notes

1 For the documents, see https://www.state.gov/the-abraham-accords/.

2 Geopolitics of the Datasphere; geode.science.

3 RIRs are five organizations that bring together Internet operators in different regions of the world. They serve different regulatory and governance purposes, and are in charge of the distribution of ASes and allocation of IP addresses. They are themselves allocated Internet resources by the Internet Assigned Number Authority (IANA).

4 For Routeviews, see http://www.routeviews.org/routeviews (accessed on 3 October 2022).

5 The Réseaux IP Européens Network Coordination Center (RIPE NCC) is one of five Regional Internet Registries (RIRs) providing Internet resource allocations, registration services, and co-ordination activities that support the operation of the Internet globally for Europe, the Middle East, and parts of Central Asia. For the Routing Information Service (RIS), see RIPEat https://www.ripe.net/analyse/Internet-measurements/routing-information-service-ris.

6 As aforementioned, the country of registration does not necessarily correspond to a geographical reality. ASes are first and foremost a set of machines that are diversely scattered across space, sometimes across different countries. When inconsistencies or suspicious interconnections are observed, we use other third-party data such as information from publicly available databases on the AS’s company’s website to inspect whether the country of registration is accurate.

7 For instance, the small Emirati ASes in Iraq are mostly satellites links, while some others in Iran have a position not easily interpretable, which will require further qualitative research.

8 See https://bgp.he.net/AS41426#_whois (accessed on 3 October 2022).

9 According to CAIDA’s AS rank. The AS rank of a given AS is a metric designed by the Center for Applied Internet Data Analysis of San Diego University to give an indication of the importance of an AS in the worldwide connectivity landscape; https://asrank.caida.org/about#rank (accessed on 3 October 2022).

10 In March 2021, the Chinese company Tencent announced plans to build a data centre in Bahrain, https://www.datacenterdynamics.com/en/news/tencent-announces-plans-first-data-center-bahrain/; and in March 2022, Bahrain announced the launch of the first data centre park in partnership with the company STC Bahrain, https://www.arabnews.com/node/2054306/business-economy (accessed on 3 October 2022).