4.1. Text analysis

We began our study by looking at the terms used in titles and abstracts to find research domains interested in maritime piracy. We selected the 500 most-relevant terms that appeared at least 10 times in the retained references. Using the text mining tool, we can identify connections between these words, resulting in a total of 52,383 links. This approach enables us to gain insights into the key concepts and themes prevalent in maritime piracy research, facilitating a deeper understanding of the various domains of inquiry within this field.

Using the terms and links, we create an undirected graph for further analysis. Figure 3 shows the constructed graph of terms for maritime piracy. In this part, the terms are not related to particular references. For this reason, the graph is analysed globally without making reference to specific research works.

Figure 3.

Most-used words in selected papers categorised in six clusters.

Multiple parameters allow for detecting clusters in different manners. In this case, the cluster algorithm detects six different clusters that correspond to main research topics and two geographic areas. Table 1 presents the identified clusters and representative terms that have been selected based on the sizes of the nodes and relation to the cluster.

Table 1.

Identified clusters through text analysis

Near clusters means closely studied subjects. The most representative case is the China and Africa clusters because, apart from the names of the regions, the studied problems are fairly similar but with few particularities. We can locate terms like ‘regionalism’, ‘democracy’ and ‘transition’ in between these two clusters. Since governments develop and implement laws, international law is likewise closely related to these clusters and then placed closely in the graph. The term ‘states’ can be easily identified as a bridge between these clusters. Similarly, we can remark that every cluster uses the word ‘law’ as a point of reference. For this reason, this node is near the centre of the graph.

Another cluster studies geographical areas and vessel trajectories. In this cluster, we find geographical terms such as area, ocean or island. These terms are linked with ‘jurisdiction’ and ‘UNCLOS’ that show the link with international law. As the geographical position and associated data of vessels are key elements when analysing the impact of piracy, multiple terms are related to them. It also contains more technical terms such as ‘signal’, ‘frequency’ or ‘AIS’. Works employing these words seem to have an engineering approach. This is because researchers look for methods for monitoring maritime traffic and detecting piracy activity. The Automatic Identification System (AIS) allows for following vessels. As pirates deactivate or modify the functioning of this system, other technical methods need to be used.

Last clusters include terms about economy and economics showing a gap between them that proves the difference between these two domains. In other words, economics focuses on population implications, whereas economy clusters concentrate on issues related to the macroeconomy or markets from a state perspective. The ‘Economics’ term is, as a consequence, connected with terms related to people like ‘sociology’ and ‘public’. In an ‘economy’ cluster, we find ‘commerce’, ‘crisis’, ‘company’, ‘capitalism’ and different names of countries. ‘Industry’ and ‘market’ appear as a bridge between these two clusters.

Numerous names of countries appear everywhere in the graph. Obviously, they are all connected with the cluster related to geographic data, but each country has different concerns and, as a consequence, different research interests. For this, each node of the graph representing countries appears in the cluster that represents their most important concerns. For example, some countries are related to the economy, while others study terrorist activities.

In addition, some themes can be discovered through these terms, although they are not identified as a cluster. Geopolitics is an example with terms such as ‘alliance’, ‘foreign affairs’, ‘dialogue’ and the names of countries.

4.2. Journals

A second analysis identifies where selected articles have been published and the references among them. The resulting graph, which displays the most significant journals in this field, is shown in Figure 4.

Figure 4.

Most-cited journals and co-citations.

In this figure, we can perceive nearly identical clusters as we did in the text analysis. That occurs because journals are very specialised. For this, every author associated with a research field prefers publishing in journals that are most suited to their work.

The most pre-eminent journal is ‘Marine Policy’ published by Elsevier. It can be found in the centre of the image and is connected to nearly every other journal. It also appears with a bigger size than other graph nodes because it contains more publications related to maritime piracy. The reason is that this publication covers general maritime topics that are likely to be of interest to other fields.

The journals on the right address regional issues. For instance, we can find ‘Asian security’, ‘The Pacific review’ and ‘Indian Ocean region’. Also, we find more general journals covering a wider range of topics in geopolitics and strategic subjects.

At the top of the image, law journals are concentrated. As perceived with the ‘law’ term in the text analysis, law journals present links within almost every journal. Two examples are ‘American Journal of International Law’ and ‘Ocean Development and International Law’. Oxford University Press’s journal ‘International Affairs’ is located at the boundary between this and the previous cluster. This is because it addresses a variety of other multidisciplinary subjects such as international governance, ethics, regional issues, politics, economy and economics.

Going to the left, we can find journals such as ‘Journal of Transportation Security’ and Cambridge University Press’s journal ‘The Journal of Navigation’ that study maritime topics in a more practical and technical manner. Lastly, the journals ‘Maritime Policy and Management’ and the ‘WMU Journal of Maritime Affairs’ establish a connection with the economics journals such as ‘Maritime Economics & Logistics’.

4.3. Articles

With the aim to identify the most known and representative articles, we perform a citation analysis. This analysis also allows us to see connections and identify important articles for particular domains or for the general topic of maritime piracy.

For citation analysis, a criterion of at least 20 citations has been adopted to ensure that articles with notable impact are considered, while still capturing fewer quoted articles for trend discovery. Following citation calculations, a selection of the 200 most pertinent references has been made to construct the graph. Among these, 164 interconnected references have been plotted in Figure 5, highlighting the network of the most influential works within the domain.

Figure 5.

Citation analysis for research articles.

First, we can easily identify two references’ groups at the top and on the right that are rather far apart from the other graph nodes. The cluster on the top is composed of works addressing alternative routes such as the Cape route (Notteboom, Reference Notteboom2012) or Arctic navigation (Theocharis et al., Reference Theocharis, Pettit, Rodrigues and Haider2018). Arctic navigation has economic ramifications since it avoids treacherous piracy routes (Zhu et al., Reference Zhu, Fu, Ng, Luo and Ge2018). As an illustration, insurance companies may agree with a piracy premium of 40$/TEU (Theocharis et al., Reference Theocharis, Rodrigues, Pettit and Haider2019). These works can be tailored to specific types of vessels, such as containers (Notteboom, Reference Notteboom2012) or dry bulk carriers (Shibasaki et al., Reference Shibasaki, Azuma, Yoshida, Teranishi and Abe2017).

Works on the right focus on the history of maritime navigation where piracy plays a significant role. There exist works concentrating on various regions and periods such as Chinese waters in the 17th century (Andrade, Reference Andrade2004), the Western Indian Ocean and Persian Gulf Regions in the 18th century (Risso, Reference Risso2001), or the western Indian Ocean from the 12th to 16th centuries (Prange, Reference Prange2011).

Technical works are shown in blue in the lower left. The most cited article looks for anomalies in AIS data (Pallotta et al., Reference Pallotta, Vespe and Bryan2013). AIS data analysis is commonly employed to study ship trajectories and knowledge extraction. In addition, maritime satellite surveillance allows worldwide vessel tracking (Fournier et al., Reference Fournier, Hilliard, Rezaee and Pelot2018). Approaching the graph’s core, we discover technical papers whose authors model the maritime piracy situation with Bayesian networks (Dabrowski and De Villiers, Reference Dabrowski and De Villiers2015). These papers had a variety of goals such as estimating the likelihood of a ship being hijacked (Pristrom et al., Reference Pristrom, Yang, Wang and Yan2016). Some of them propose agent-based systems with different purposes such as enhancing transport security (Jakob et al., Reference Jakob, Vaněk and Pěchouček2011) or search operations (Onggo and Karatas, Reference Onggo and Karatas2016).

At the bottom, in red, maritime piracy is examined as a geopolitical issue (Germond, Reference Germond2015). Several research articles make links between Illegal, Unreported and Unregulated (IUU) fishing activities and piracy (Okafor-Yarwood, Reference Okafor-Yarwood2019) or other related conflicts (Pomeroy et al., Reference Pomeroy, Parks, Mrakovcich and LaMonica2016). An important part of these works focuses on national and transnational measures to address maritime security (Chapsos and Malcolm, Reference Chapsos and Malcolm2017). For instance, some European vessels carry weapons aboard to repel pirate attacks, so they can no longer legally dock in ports where weapons are prohibited (Le Manach et al., Reference Le Manach, Andriamahefazafy, Harper, Harris, Hosch, Lange, Zeller and Sumaila2013).

Law and policy articles mixed with works analysing geopolitics, which are at the heart of the graph, serve as a bridge between various academic domains. This mix is due to the proximity of these topics because policy concerns the rules or plans while politics is the process to reach them. These articles typically use state examples, most notably Somalia (Treves, Reference Treves2009). As indicated before, some works make reference to geopolitical concerns. For example, the sophistication of piracy attacks may be related to weak or failed states (Hastings, Reference Hastings2009).

This graph allows confirmation that even if most of the clusters are connected, there exist big gaps between some domains. For instance, as stated before, teams working on geopolitics are quite connected with works concerning law, while they are less connected with technical works.

4.4. Authors

To produce a clearer visualisation, we have decided to reduce the number of retained authors. For this purpose, we selected the most cited ones having at least five works in the domain. After, we keep authors who are linked to other retained authors. This allows preserving authors that belong to the ‘maritime piracy research community’. For this analysis, we have tested with different filter parameters, obtaining similar results. In this manner, a shortlist of 33 authors is produced. The resulting graph is presented in Figure 6. In the figure, the yellow cluster (top right) has been shifted to reduce the image size since it was located far to the right.

Figure 6.

Most-cited authors related by citations.

We can easily perceive that authors are connected by groups creating clusters. These relations are obviously related to the research articles’ citations. As a result, clusters are nearly the same as in previous subsections, but now, we can identify significant authors. In the next paragraphs, we present the most-cited authors who are represented in the figure by being more interconnected and having bigger nodes.

Bueger (Reference Bueger2015) (bottom left in orange) is a researcher in international political relations and is one of the more cited authors in this field. His works deal with maritime security and more precisely with piracy (Bueger, Reference Bueger2015). Because of his broad perspective, he has become a frequently cited source.

Edmunds, a co-author of Bueger et al. (Reference Bueger, Edmunds and McCabe2020), believes that piracy is a challenge to include in the maritime agenda. On the left, in blue, Germond (Reference Germond2015) illustrates the geopolitical aspect of maritime piracy.

In the middle of the graph, Daxecker and Prins (Reference Daxecker and Prins2013) worked together on the resolution of piracy as they argue that piracy is a result of permissive institutional environments and the lack of legal forms of employment in fishing sectors. Percy and Shortland (Reference Percy and Shortland2013) have studied a similar topic, concentrating on pirates from Somalia. As mentioned in earlier sections, J.V. Hastings is renowned in geopolitics for his work establishing relations between state issues and piracy (Hastings, Reference Hastings2009).

At the bottom of the graph, in red, Brewster (Reference Brewster2017) is an expert in the Indian Ocean navigation with a geopolitical and strategic point of view; similarly to Bateman (Reference Bateman2010). Some writers in this cluster, like C. Liss, focus more on piracy issues as a transnational issue (Liss and Sharman, Reference Liss and Sharman2015). Adolf K.Y. Ng. links this cluster with other research topics studying the impacts of piracy on economics (Fu et al., Reference Fu, Ng and Lau2010).

In the top green cluster, Yang appears to be a key node working on models to protect merchant vessels (Pristrom et al., Reference Pristrom, Yang, Wang and Yan2016) and spatio-temporal patterns (Li and Yang, Reference Li and Yang2023). Thanks to their contributions to agent-based modelling, Vaněk et al. (Reference Vaněk, Jakob, Hrstka and Pěchouček2013) appear at the top of this cluster.

On the upright, in yellow, an Italian team of researchers using tracking techniques for vessel surveillance and searching for counter-piracy goals is depicted. Vespe et al. (Reference Vespe, Greidanus and Alvarez2015) makes the link with the main graph with works using AIS and LRIT (Long Range Identification and Tracking) data. This cluster demonstrates that there exist research groups that are linked by proximity in terms of laboratory location and topics. By adjusting the analytic parameters, we could find Pallotta et al. (Reference Pallotta, Vespe and Bryan2013) within this cluster. She shares works with the other authors and had yet to be identified in previous analysis. Since this author is a valuable resource for AIS data analysis and anomaly detection, but not for the field of maritime piracy, her graph node disappears with more restrictive filters.

4.5. Identified research topics

Thanks to the presented analysis, we have identified seven main types of topics. They have been initially identified thanks to text analysis, and reinforced by journal and citation analysis. History has been identified afterwards because it focuses on the same topics but with a historic vision. Table 2 lists them with an associated definition.

Table 2.

Main identified research topics

Table 2 shows the diversity of research topics pertaining to maritime piracy. This demonstrates how challenging it is to approach this issue from a single perspective. With this vision, we can affirm that maritime piracy is a historical problem that is concentrated in specific areas. It has multiple origins, usually linked to economy and economic issues. The solutions are numerous, heterogeneous and proposed from different research domains such as law, economy and technology.

4.6. Current trends

To detect last years’ trends, we carry out a text mining analysis for publications written between 2019 and 2024. This time, no citation limit is imposed. Some of these works did not have enough time to be cited because they are recent and did not appear in previous analyses. The same filtering thesaurus is applied as introduced in Section 3. To simplify the results, we chose to retain terms cited at least 40 times. The result is depicted in Figure 7.

Figure 7.

Most-used words in research publications between 2019 and March 2024.

Some statements can be easily perceived. The Africa and China regions are still of important interest. These works follow the piracy evolution through time. Economy and economic issues related to maritime piracy are now blurred and more difficult to identify. International law is always crucial when addressing piracy, and it can be easily identified at the bottom of the graph. That is because States should always converge diverse points of view into a shared position based on the context evolution. Seafarers are constantly looking for new, better, more efficient and safer routes; this is referred to as ‘performance’, which is primarily associated with ‘risk’ and ‘time’.

Concerning ‘China’, we can appreciate that now it is more connected to other important nodes such as ‘India’ and ‘Asia’. It proves the importance of the SEA (South-East Asia) region in contemporary maritime piracy hotspots. ‘Cooperation’ and ‘strategy’ are near terms that show the collaboration between the countries located near this zone.

New terms, such as ‘pandemic’ and ‘covid’, indicate new trends based on current events. A significant transformation is expressed by the term ‘data’. It appears in the middle of the red cluster and it is related to numerous fields. We have also perceived that this node also grows over time. New information technologies offer new possibilities for pirates and to develop new protection systems. This is the main reason of Section 5 describing the mainly used datasets. Additionally, the word ‘History’ is now clearly visible on the right and it is mainly connected with ‘Singapore’ due to the significance of the Malacca Strait in global maritime trade.

Although it has not been identified as a cluster, a group of nodes that is easily identifiable, between the red and green clusters, is composed of terms like ‘port’, ‘time’, ‘vessel’, ‘performance’, ‘course’ and ‘ship’. They show the increasing number of works analysing the movements of vessels. This is mainly due to new technologies such as global AIS or satellite images.

As suggested in the introduction, cybersecurity can be considered a new era of piracy in the maritime domain. Recent cyberattacks impacted important actors such as Maersk in 2017, CMA-CGM in 2020 or the Barcelona port in 2018. These events show the present and actual risk, creating important operational and economic consequences. Cyber risks have been recently identified by the IMO with Resolution MSC.428(98) as an important risk to consider when applying ISM and ISPS Codes. Because of this, cybersecurity researchers look to apply their knowledge to the maritime sector.

5.1. Datasets concerning maritime piracy acts

Historical works, evolution of attacks, or time or geographical analysis are based on historical datasets. When analysing presented works using data, we can identify three datasets that are usually used.

The first corresponds to data collected from the International Maritime Bureau (IMB), cleaned and enriched with geospatial data. This dataset contains information on maritime pirate attacks between January 1993 and December 2020, as well as country indicator data for the same period (Benden et al., Reference Benden, Feng, Howell and Dalla Riva2021). It has been used by several works with different objectives, such as validating hypotheses about maritime piracy (Daxecker and Prins, Reference Daxecker and Prins2013) or detecting patterns improving these data (Coggins, Reference Coggins2012). Figure 8 illustrates the various pieces of information provided by the dataset.

Figure 8.

IMB database schema for piracy acts (Benden et al., Reference Benden, Feng, Howell and Dalla Riva2021).

Second, Piracy and Armed Robbery (PAR) dataset that is associated with the Global Integrated Shipping Information System (GISIS) has been collected by the International Maritime Organization (IMO) and is freely available on their website.Footnote ⁶ The PAR dataset covers worldwide pirate attacks from January 2006 to the present day. Piracy incidents are listed with numerous characteristics, including location, date, incident details and incident consequences. For illustration, this database has been used to estimate the likelihood of a ship being hijacked in specific areas (Pristrom et al., Reference Pristrom, Yang, Wang and Yan2016). Table 3 shows the different information provided by the dataset.

Table 3.

Description of PAR dataset extracted from GISIS

Finally, the third dataset corresponds to the ASAM (Anti-Shipping Activity Messages) piracy dataset. It is available on its website,Footnote ⁷ and provides information on the type of vessel targeted, date, position, as well as a description of the attack. This dataset has been used, for example, to analyse the dialectics of piracy in Somalia (Samatar et al., Reference Samatar, Lindberg and Mahayni2010). Table 4 shows the different provided information.

Table 4.

Description of the ASAM dataset

These three databases are different and do not contain the same information. As a result, some authors have chosen to use the most relevant one for their analysis or to merge them in different manners. For instance, the second and third datasets have been merged to find relations between fishing issues and maritime piracy (Desai and Shambaugh, Reference Desai and Shambaugh2021). Additionally, these three datasets can be cross-referenced and joined to create a new, more comprehensive and complete dataset (Li and Yang, Reference Li and Yang2023).

5.2. Datasets concerning AIS data

Nowadays, ships are predominantly observable in real-time using various position sensors, aiming to identify and locate these vessels and to detect anomalies at sea. The surveillance of maritime areas, especially coastal regions, for safety and security reasons, traffic management or protection of biodiversity zones, relies heavily on identifying typical ship positions and behaviours that enable the detection of anomalies or abnormal trajectories. This is done to detect illegal or criminal activities, associated risks (such as piracy, illicit trafficking, clandestine immigration, over-exploitation of fishery resources, pollution from hazardous materials, accidents, etc.) and violations of regulations. The multitude of ship positioning reporting systems implemented generates constantly increasing volumes of data to analyse. For this, numerous research works addressing maritime piracy are interested in these data (Pallotta et al., Reference Pallotta, Vespe and Bryan2013; Vespe et al., Reference Vespe, Greidanus and Alvarez2015).

The AIS is a system that has been on board ships for almost two decades. Originally designed as an anti-collision system for ships, it is mandatory for all ships subject to the SOLAS convention; ships over 300 gross tonnage, fishing vessels over 15 meters and ships carrying passengers. The system essentially consists of a GNSS (Global Navigation Satellite System) as GPS (Global Positioning System), a transponder and a VHF (Very High Frequency) antenna for communication. The data exchanged by a ship equipped with the AIS includes static information (ship name, international identifier, size, age, cargo, etc.) and dynamic information (heading, speed, GNSS position, etc.). The system also transmits additional meta-information related to the trip (destination, expected arrival date and time) as well as control messages. Recently, AIS data have gained much popularity in maritime research (Yang et al., Reference Yang, Wu, Wang, Jia and Li2019). In our bibliometric dataset, 205 research articles use this data directly. 447 articles appear when no citation restriction is applied. Some research works look for enhancing SAR systems (Zhao et al., Reference Zhao, Ji, Xing, Zou and Zhou2014), detecting anomalies (Kontopoulos et al., Reference Kontopoulos, Chatzikokolakis, Zissis, Tserpes and Spiliopoulos2020; Pallotta et al., Reference Pallotta, Vespe and Bryan2013) or improving awareness in specific zones (Posada et al., Reference Posada, Greidanus, Alvarez, Vespe, Cokacar and Falchetti2011).

A non-exhaustive list of sources concerning AIS data is as follows.

• MarineTraffic https://www.marinetraffic.com/ is a leading maritime analytics provider, offering real-time insights into ship movements, harbour activities and port locations. Through its platform, users gain access to a comprehensive database containing vital vessel information such as construction origin, dimensions, gross tonnage and International Maritime Organization (IMO) numbers. While accessing vessel positions is freely available through the MarineTraffic website, the comprehensive data needed for in-depth analysis is not provided for free. Accessing and storing data for analytical purposes typically involves subscription or payment for premium services. Figure 9 shows an example of these data.

  Figure 9.

  Example of MarineTraffic data in the Arabian Sea.

  

• Marine cadastre https://marinecadastre.gov/ offers free access to AIS data, which provides real-time and historical information on vessel movements in U.S. waters. MarineCadastre.gov also facilitates access to a variety of spatial data sets and tools pertaining to the marine environment along the American coasts. These data can include information on marine boundaries, marine protected areas, marine habitats and marine energy resources.

• Other academic datasets provide some methods to analyse or store data (SQL scripts) as well as supplementary data (fishing areas, marine protected areas, ports), such as Ray et al. (Reference Ray, Dréo, Camossi, Jousselme and Iphar2019), where data involved concern Brittany coasts. Another dataset concerns Greece coasts with supplementary weather data and extraction of critical points for synthesised trajectories (Tritsarolis et al., Reference Tritsarolis, Kontoulis and Theodoridis2022).

The last two AIS datasets are less used than the first one. This is because the first covers the entire world, whereas the others concentrate on particular areas. However, they offer insightful viewpoints for this field and are frequently employed to test novel anomaly detection techniques.

AIS data present some limitations. As this system has only been an OMI requirement since 2000, it is difficult to realise pertinent analyses before this year. Also, transponders can be easily switched off or modified, and they are subject to GNSS issues (Banaei Kashani, Zhang, Hendawi, ). As a consequence, analyses based on these data need to take into account these limitations. It should be noted that researchers are looking for more secured systems (Goudossis and Katsikas, Reference Goudossis and Katsikas2019).