My current PhD research focuses on the use of machine learning techniques and qualitative research methods in identifying online child sex trafficking patterns in Southeast Asia. This is an under-researched area, yet important and closely linked to the context of UK, US, and Europe. As the organized crime sector is growing on a transnational scale, cross-sectional and interdisciplinary research is becoming more important. According to the Internet Watch Foundation (2019), 95% of the world’s child sexual abuse material is hosted in Europe and North America, and more than half of the victims (73%) are in Asia and the Pacific. With my experiences of working with NGOs, researchers, and law enforcement officials in the field of Security and Intelligence in Vietnam, Cambodia, and Uganda, the research aims to: (1) Explore the application of data science/computer science in studying criminal activities, thus develop early detection tools that assist in the investigation and victims rescue process; (2) Build capacity, strengthen legal framework and collaboration across disciplines (academia, NGOs, government) and countries (UK, Vietnam, and Cambodia) through roundtable discussions and technical support; and (3) Provide human rights approaches and NGOs perspective in promoting effective victim protection and assistance. As the crime is trans-national (e.g., involving demands from Westerners, illicit webs hosted abroad, etc.), the research has the potential to be replicated in other countries where child sex trafficking is prevalent and assist in uncovering the international criminal networks involving child-sex exploitation.
The rapid growth of digital technology in recent decades has expanded the opportunities for cultural and economic exchange in a globalised world. However, the increase in interconnectedness has come accompanied by an unprecedented surge of illegal exploitation and misuse of technology for criminal activities (Sarkar, 2015). Sex traffickers and organised criminal groups have been using the Internet and other telecommunication technologies to recruit, market, and deliver women and children into sexual exploitation and modern-day slavery (UNODC, 2015). Though sex trafficking represents only 20% of global human trafficking victims, it accounts for 66% of the profits, making it a highly lucrative industry (Human Rights First, 2017). Globally, more than 1 million children are trafficked for sex annually (ILO, 2017). Remarkably, 95% of the world’s child sexual abuse materials (CSAM) are hosted in Europe and North America, with the Netherlands alone hosting 47% (The Internet Watch Foundation, 2019). Of the 4.8 million victims of forced sexual exploitation globally, about 73% are in Asia and the Pacific (Global Slavery Index, 2018). In Southeast Asia (SEA), due to the consolidation of policies and law enforcement against the sexual exploitation of children in Thailand and the Philippines, the human traffickers operators has been diverting to Vietnam and Cambodia (Ecpat, 2018).
Child-sex trafficking defined as the recruitment, harboring, transportation, provision, obtaining, or advertising of a minor child for the purpose of a commercial sex act, which involves the exchange of anything of value – such as money, drugs or a place to stay – for sexual activity. The rise in child trafficking in the Southeast Asia is linked to the alarming increase in online child pornography, including live streaming of sexual abuse of children. Despite the documented increase in the online child sex trafficking, little is known yet about child-sex trafficking networks and how transactions in markets for child-sex take place. This study is the first attempt to use data mining and machine learning techniques combined with qualitative research methods to develop analytical tools that allow the early detection and prevention of child-sex trafficking, as well as, to uncover the structure and functioning of child trafficking networks in the context Vietnam and Cambodia, two Southeast Asian countries where the problem of child-sex trafficking is particularly acute.
Specifically, the research aims to: (1) compile a corpus of data on child sex trafficking activities in Vietnam and Cambodia using data mining and text analysis techniques and online data from social media channels used for child-sex exploitation; (2) enrich our understanding of victimization patterns and experiences of victims using qualitative research methods (e.g., observations, interviews, focus groups); and (3) combine quantitative and qualitative data from (1) and (2) to build predictive algorithms for detecting online child sexual trafficking and support early interventions and the prevention of illegal activities involving children.
Importantly, insights into child exploitation and the analytical tools designed to identify and prevent those illegal activities will be of application to other SEA countries such as Thailand and Philippines where online child sexual exploitation is prevalent. The research will benefit and have an impact on those countries by: (1) enhancing the capacity of law enforcement and NGOs through training on new techniques for detecting online child trafficking and through social network analysis of offenders/victims; (2) strengthening the collaboration across the disciplines (academia, NGOs, Government) and countries (UK, Vietnam, and Cambodia) through roundtable discussions and technical support; (3) producing evidences and data will be used to make informed decisions on policies and strengthen the legal framework in those mentioned countries. As the crime is trans-national (e.g., involving demands from Westerners, illicit webs hosted abroad, etc.), the research has the potential to promote the prevention and investigation of international criminal network involving child-sex exploitation.
2. Literature review
Despite the rapid growth of child-sex trafficking activities in SEA countries in recent decades, there is still a significant lack of data related to those activities in those countries. (Szekely et al., 2015; Sarkar, 2015; Merdian et al., 2019). As most studies on sex trafficking are conducted using survey techniques and individual interviews, there is a significant lack of quantitative data related to human trafficking (Szekely et al., 2015; Sarkar, 2015; Merdian et al., 2019). While sample-based approaches are useful in some contexts, it is difficult to perform large-scale comparative analyses stratifying by geography, ethnicity, or class of services (Dubrawski, 2015). Additionally, due to the sensitivity nature of the issue, the under-reporting of cases undermines the real magnitude of the issue (Dubrawski, 2015).
In contrast with survey data, open-source data from the Internet offers an incredibly rich source of information on child exploitation activities (location, frequency, telephone number, etc.), as well the profiles of offenders and victims. In recent years, there has been important advances in the use of data mining and machine learning techniques to detect human trafficking patterns online (Wang et al., 2012; Silva et al., 2014). For example, recent contribution proposed the use of image and textual analysis (Alvari et al., 2016; Latonero, 2011), sentiment analysis (Anastasija et al., 2018), and social network analysis (Cockbain et al., 2011) to identify patterns and allow the early detection of child-sex trafficking and pave the way for the application of data analytics and machine learning in assisting the investigation and victims rescue. As the tasks of manually sorting is a strain on the limited human resources of law enforcement agencies, data mining and machine learning appear as technological solutions to be used with proprietary data to uncover cases of human trafficking (Marturana et al., 2013; Kena et al., 2014) and to gain insights into the nature of such activity that may facilitate law enforcement. However, firstly, most of the research to date has studied the US or UK contexts using online data from platforms such as backpage.com, craiglist.com, and myspace.com (Wang et al., 2012; Silva et al., 2014; Alvari et al., 2016; Latonero, 2011). No equivalent research has been conducted in Southeast Asian countries where child-sex trafficking is most prevalent and most demands for those illegal activities from the US, UK, and Europe are met. Findings and analytical tools developed for the US and UK are not directly applicable to countries in Southeast Asia. Secondly, languages used to build and test the algorithms are in English, but local languages – the most common languages used to exchange information on recruitment, selling, and marketing are missed. Thus, the identification of patterns in illegal activities in those countries requires different text analyses and algorithms. Thirdly, space-time clustering analysis – an emerging field to study epidemics, crime, and terrorist activities has immense potentials for anti-trafficking studies, but has not yet been applied.
Parallel to the quantitative research aimed to detect child-sex trafficking activities, there is a body of qualitative research aimed to understand the context and circumstances surrounding those activities. Most of that research to date, however, has focused on the experiences of victims (Blackburn et al., 2010), legal frameworks (Ecpat, 2018), health-psychological consequences (Beazley, 2014); or social determinants (Davy, 2013) of offline child-sex trafficking. Yet little is known yet about online offenders and their networks, the grooming tactics used to engage with children, and platforms used to carry out online child sex trafficking in Southeast Asian contexts. This project aims to fill this gap by using materials from real cases of child-sex trafficking through the interviews with the law enforcement officials and victims to improve our understanding of the operation of activities in those countries. This research, thus, is an interdisciplinary study which will combine ethnographic research, data analytics and machine learning to provide an updated perspective and propose interventions for child sex trafficking in Southeast Asian countries.
3. Proposed project: Aim, research questions, and research objectives
Aim: This study aims to contribute to the literature on child-sex trafficking by exploring the use of data mining and machine learning techniques to understand linguistic, demographic, and geographic patterns in child-sex trafficking activities in Southeast Asia. This will be combined with qualitative techniques to understand the functioning of child-sex trafficking networks and the circumstances of victims and offenders involved in those networks.
- How to build a corpus with relevant information on human trafficking/child sex trafficking reflecting English and local languages used in SEA? What trafficking behavioural patterns can be identified?
- What are the characteristics of offender and victim networks in SEA? How can these features be used to understand the crime networks?
- How different are trafficking maps during major events?
- Is it possible to build a predictive model of human trafficking/child sex trafficking in SEA?
1. To improve our understanding of sex-trafficking networks and the circumstances surrounding the lives of those involved in those networks including victims and offenders.
2. To build a corpus of online advertisements in English and SEA languages related to child sexual exploitation that could be used by analysts to build and train machine learning algorithms to detect sex-child trafficking activities.
3. To combine the corpus of online adverts created in 1) and the qualitative data generated to address goal 2) to build a machine learning algorithm to uncover temporal, geographic, and demographic patterns in child-sex trafficking activities. These algorithms will be applied to investigate patterns of child-sex trafficking in the context of the 32nd SEA games, which will take place in Cambodia, 2023.
4. Theoretical framework of the applications of data analytics and machine learning on child sex trafficking
A combination of computer-assisted analysis and data collection with a human expert making informed decisions regarding that data could increase the likelihood of detecting possible cases of child sex trafficking online. The use of analytical approaches such as image analytics, textual and visual analytics, predictive and network analytics, and web-based tools can enhance existing intelligence based investigations of law enforcement agencies (LEAs) (Konrad et al., 2016). Combining open-source data analytics, ontological knowledge representation and a wider notion of knowledge management can provide LEAs with useful information to identify, track, and prosecute traffickers, as well as potentially rescue and protect the victims.
- Textual analysis: Text analytics encompasses a variety of semantic and linguistic disciplines such as sentiment analysis, data mining, concept and contextual extraction, and content categorization (Hepburn et al., 2010). Social media mining is one of the emerging fields in quantitative anti-trafficking studies (Latonero et al., 2011; Dubrawski et al., 2015). Social networking, chat, dating and community websites such as Facebook, Twitter, Tinder, Instagram, and public escort service ads contain abundant and under-utilised traces of human trafficking activity (Dubrawski et al., 2015; Territo et al., 2009). Information such as ages, locations, prices of services, posting dates, and images can be leveraged to quantify the prevalence of trafficking and to characterise the populations involved and their modes of operation. Maps of trafficking locations or high-activity areas could help to identify trafficking patterns and recurrences of trafficking-related activities. One of the challenges is to develop techniques to detect child sex trafficking signals while filtering the vast information of noise from adult escort services (Latonero et al., 2011). Additionally, there are increasing deliberate efforts in data obfuscation, such as non-random misspellings of common words, high occurrences of out-of-vocabulary words, and frequent use of Unicode characters, making natural language processing (NLP) techniques a more challenging task (Brewster et al., 2014).
- Social network analysis (SNA): A network is a set of entities, usually individuals, organizations, or commodities, connected by links which symbolize relationships and interactions (i2, 2010). SNA can reveal the relative importance of each entity by analysing power as conferred by links to other network members (Hanneman et al., 2005). In human trafficking intervention, this can reveal key relationships within criminal structures as the best chance for police to disrupt or disintegrate the trafficking network. Or social networks of at-risk persons can be used to determine which contacts have critical influence over others, thus enabling early identification of victims, or those that have the highest likelihood to propagate the information flow (Cockbain et al., 2011). Victim and offender networks using in-built centrality metrics (degree, closeness, and betweenness) will be constructed to measure a network’s overarching structural properties and identify powerful individuals.
- Space-time clustering analysis: Spatial autocorrelation and space-time clustering analysis have been used to study epidemics (Knorr-Held et al., 2003), crime (Eck et al., 2005; Johnson et al., 2004); and terrorist activities (Gao et al., 2013). In human trafficking, traffickers take advantage of increased demand for commercial sexual exploitation during major sport events and conventions (Austin, 2011). Analyses of Big Bowl events in the US represent a showcase for such study (Szekely et al., 2015; Brewster, 2014; Hovy et al., 2014). The transportation of victims surrounding such events presents an opportunity to gain understanding and predict temporal and spatial aspects of the problems. The 2023 Southeast Asian Games (known as 32nd SEA Games) which will be organised in Cambodia, presents an opportunity to study such patterns.
5. Proposed Design and Methodology
This is a three-year research project in which research activities will be structured in three phases. The first phase will focus on the construction of the corpus that will be used to develop and train the algorithm to identify child-sex trafficking activities. For this, we will use text analysis techniques to detect suspicious child sex trafficking activities in the marketing of escort services in online social media channels such as craiglists.com, expatforum.com, and expat-blog.com..The second phase will make use of qualitative data from interviews with law enforcement officers and victims in Vietnam and Cambodia to investigate child-sex trafficking practices and the circumstances surrounding the different parts involved in those practices. In case the data is available, social network analysis technique is applied to understand to network patterns of the offenders and victims, thus, offers an investigative tool for breaking the network. In the third phase, quantitative and qualitative data from phases 1 and 2 will be combined to develop and train a predictive model to enable early detection and prevention of child-sex trafficking. The model will be applied to study the extent and patterns of child trafficking in the 32nd SEA Games that will be held in Cambodia in 2023.
In what follows I describe in detail each phase of the project and how they will contribute to the goals of the project.
5.1. Building datasets of child sex trafficking through data mining, processing and algorithm training (phase 1)
This phase will commence with the student conducting a review of the academic and grey literature and other materials related to child sexual trafficking in Southeast Asia, with a particular focus on Vietnam and Cambodia). The review will consider a range of sources including journal articles, policy reports, newspapers, social media, illicit web domains, and forums containing relevant information child sex trafficking activities. The purpose of the literature reviews is to gain a deep and most up-to-date understanding of the child-sex markets in the region.
For the construction of the data corpus, online escort services advertisements will be extracted from various open websites such as craiglist, expatforum, and expat-blog.com, etc. using crawling techniques. The information extracted from online advertisements may include image and video but, given the scope of this research, I only focus on the textual component of the data. Once the data have been extracted, the dataset (including 500 advertisements – for training and testing purposes) will be sent to the law enforcement agency for classifying suspicious cases of child sexual trafficking. A dictionary of frequent keywords, terms of interests, expressions in English/local languages that might indicate human trafficking/sexual exploitation will be compiled. As most of the trafficking victims have low literacy or are unable to speak English, and their communication is through local languages, identifying slang and code-words is essential in understanding trafficking situations at the local level (Ecpat UK, 2019). Experiences from the US show that the law enforcement officials routinely look for cues within online escort services that might indicate child trafficking situation. For examples, the cues may include the use of a third-person voice, certain keywords (cuties, schoolgirls, babies, etc.). By applying unsupervised or semi-supervised learning techniques, the machine learning models will be trained to partially automate identifying instances of child trafficking. Several machine learning algorithms such as Conditional Random Fields (CRF), Word Feature Vectors (WFV), Support Vector Machines (SVM) will be trained and evaluated by comparing the precision, recall with the testing dataset. The algorithms with the highest precision will be selected as a predictive model to identify suspicious cases of child sex trafficking. This allow the monitoring of all newly observed escort ads and make alarms to the law enforcement officials for early detection and investigation.
Additionally, multi-dimensional data sets which combine time, location, phone usage, prices, nature of ads will be constructed. By analyzing the information, patterns such as (1) characteristics of potential traffickers (pimp or madam), clients of the trade (johns), and the victims; (2) possible routes/trade; (3) mapping of hotspots can be complied. The analysis of this dataset can help contribute to the understandings the patterns of the sex trafficking networks and the magnitude of the problems in the mentioned countries.
5.2. Constructing offender and victim social networks and constructing narratives of child sex trafficking (phase 2)
Based on the availability of data, I will employ social network analysis and/or qualitative research to gain insights into the child sex trafficking network and patterns.
a. Sex trafficking social network analysis
To answer question 2, based on interviews with 30 offenders and 15 victims, and 20 law enforcement officials across three countries, case summaries, and court visits, victim and offender networks will be constructed. Each network will be drawn based on three centrality metrics including degree, closeness, and betweenness (Cockbain et al., 2011). Each individual entity’s scores from 0 (low) to 100 (high) and descriptive statistics (mean, range, etc.) for each network will be calculated.
Once the network is built, it can be used to understand how the victims were targeted, the mode of operation of the offender network (collectively or individually), and the potential that victims might become future pimps or recruiters, or key holders in the offenders network, thus shedding light on the current operations of trafficking network (Cockbain et al., 2011). Though new material can emerge and metrics may need to be re-run, social network analysis can help counter personal biases and suggest appropriate interventions.
b. Qualitative research on child sex trafficking networks
For this phase, I will conduct 20 interviews with the law enforcement officials and 30 interviews with victims in each country (Vietnam and Cambodia). Participants will be recruited in collaboration with a series of institutions involved in the fight against child sexual trafficking including the Ministry of Public Security (Vietnam), Action Pour Les Enfants (Cambodia), Anti-human Trafficking Unit of the Cambodia Law Enforcement Officials (Cambodia) and a network of NGOs (such as Pacific Foundations, Terre des Hommes, Ecpat, etc. in Vietnam and Cambodia) with which the student has a history of collaboration. As for the victims, an additional survey for 100 girls/boys in the Safe shelters in Vietnam and Cambodia (50 each) will be carried to learn about the background and socioeconomic circumstances of victims, as well as, to identify common platforms that the offenders use to access the children, timespan connection between offenders and victims, and other techniques that the offenders used to groom the children. These data will be used to gain insight into the child-sex networks and their activities. In addition, all qualitative data gathered in this phase of the project will be added to the data corpus created in Phase 1 that will be used to develop and train the predictive algorithm in Phase 3.
5.3. Testing predictive model of child sex trafficking algorithm to the 32nd SEA games (phase 3)
Based on the quantitative and qualitative data collected during years 1 and 2, inputs for algorithms will be modified and outputs of each algorithm will be evaluated using the precision, recall, and F1-score measurements (Burbano et al., 2017) (question 3 and 4). Algorithms with best performance will be selected and used to construct a live map of the 32nd SEA Games of possible child sex trafficking. A temporal distribution of escort will be designed to compare the rate of trafficking/crime during the events and normal time (one month before and after the event). A spatial map based on the geo-spatial location of the ads is used to construct the crime maps, including usage of dots, points, line, and shaded areas representing individuals, streets, and area levels of potential trafficking. Police responses during the events are interesting factors to test different crime theories (routine activity theory, offender search theory, disorganization theory, opportunities theory, etc.) and to reconstruct crime maps over space and time. The analysis will contribute to the literature of human trafficking during major sport events and policing recommendations (Benjamin, 2007).
6. Ethical considerations and further development
The research will obtain ethical approval from the University and will comply with all the ethical requirements for this type of research. Participation of respondents are voluntary and have rights to withdraw from the study at any stage if they wish to do so. Informed written consent will be sought from all respondents. The consent involves the student providing sufficient information and assurances about taking part to allow individuals to understand the implications of participation and to reach a fully informed, considered and freely given decision about whether or not to do so, without the exercise of any pressure or coercion. Research participants are not subjected to harm in any ways whatsoever. As the research involves interaction with victims and/or their families, questionnaires will be designed in a way that do not trigger trauma to the victims, especially the children. All the questionnaires and surveys will sesek for advice from a child protection specialist before release. Privacy and anonymity or respondents is of a paramount importance. Ultimate confidentiality will be kept in regards to the individuals. The research shall adhere to the Data Protection Act and comply with national law on data protection and data privacy of the countries where the research takes place.
Findings will be shared with academics, NGOs and law enforcement stakeholders (ECPAT UK, Pacific Foundations, National Law Enforcement Officials, etc.) via cross-sector round-table events. The final work will be prepared as a comprehensive knowledge graph and user-friendly version to assist the work of NGOs and law enforcement agencies in investigation and victim protection. The results will also be shared with high impact interdisciplinary journals (e.g., Journal of Human trafficking, Journal of Artificial Intelligence Research, British Journal of Criminology) and conferences (e.g., IEEE Conference on Intelligence and Security Informatics, International Conference on World Wide Web). The research will be developed further in collaboration with current initiatives on identifying human trafficking/online child sex trafficking such as Thorn, Thompson Reuters, Mekong Club, etc.
- Alvari, H., Shakarian, P., & Snyder, J. (2016). A non-parametric learning approach to identify online human trafficking in IEEE Conference on Intelligence and Security Informatics (ISI). IEEE, pp. 133– 138.
- Anastasija, M., & Chris, A. (2018). Ensemble Sentiment Analysis to Identify Human Trafficking in Web Data. In Proceedings of ACM Workshop on Graph Techniques for Adversarial Activity Analytics (GTA 2018). ACM, New York, USA.
- Austin, L. (2012). Games and human trafficking research. Report. London Councils.
- Benjamin, P. (2007). Faster, Higher, Stronger: Preventing Human Trafficking in the 2010 Olympics. The Future Group.
- Blackburn, A.G.; Taylor, R.W.; Davis, J.E. (2010). Understanding the Complexities of Human Trafficking and Child Sexual Trafficking: The Case of Southeast Asia. Women & Criminal Justice. 20:1-2, 105-126.
- Borgatti, S. P., Mehra, A., Brass, D. and Labianca, A. (2009). Network Analysis in the Social Sciences. Science 323(5916): 892–5.
- Bracket Foundation. (2018). Artificial Intelligence: Combating online sexual abuse of children.
- Brewster, B., Polivina, S., Rankin, G., & Andrews, S. (2014). Knowledge management and human trafficking: using conceptual knowledge representation, text analytics and open-source data to combat organized crime. In: Hernandez, N., Jaschke, R. & Croitoru, M., (eds.) Graph-Based Representation and Reasoning. Lecture Notes in Computer Science (8577). Springer International Publishing, 104-117.
- Burbano, D., & Hernandez-Alverez, M. (2017). Identifying human trafficking online. in IEEE Conference on Intelligence and Security Informatics (ISI).
- Cockbain, E.; Brayley, H.; Laycock, G. (2011). Exploring internal child sex trafficking networks using social network analysis. Policing. 5, 144–157.
- Cordua, J. (2019). A Bold Goal: Eliminating Child Sexual Abuse from the Internet. Thorn.
- Davies, H.; Beddoe, C.; McCourt, S.; McIntyre, J.; Geden, J.; Morgan, A. (2013). Civil Prevention Orders Sexual Offences Act 2003: ACPO Commissioned Review of the Existing Statutory Scheme and Recommendations for Reform.
- Davy, D. (2013). Understanding the motivations and activities of transnational advocacy networks against child sex trafficking in the Mekong Sub-region: The value of cosmopolitan globalisation theory. Cosmopolitan Civil Societies Journal. 5, 39–68.
- Dubrawski, A., Miller, K., Barnes, M., Boecking, B., & Kennedy, E. (2015). Leveraging Publicly Available Data to Discern Patterns of Human-Trafficking Activity. Journal of Human Trafficking, 1(1), 65–85.
- Eck, J.; Chainey, S., Cameron, J., Leitner, M., & Wilson, R. (2005). Mapping crime: Understanding hot spots, Report, National Institute of Justice Special Report.
- Ecpat UK. (2019). Precarious journeys: Mapping vulnerabilities of victims of trafficking from Vietnam to Europe. Available online: https://www.ecpat.org.uk/precarious-journeys. Accessed 02/01/2020.
- Gao, P., Guo, D., Liao, K., Webb, J., & Cutter, S. (2013). Early detection of terrorism outbreaks using prospective space-time scan statistics. The Professional Geographer 65 (4) (2013) 676–691.
- Global Slavery Index. (2018). Factsheets: Asia and the Pacific. Available online: https://www.globalslaveryindex.org/2018/findings/regional-analysis/asia-and-the-pacific/. Accessed on 02 Jan 2020.
- Han, B., Cook, P., & Baldwin, T. (2014). Text-based Twitter user geolocation prediction. Journal of Artificial Intelligence Research. 49, 451-500.
- Hanneman, A. & Riddle, M. (2005). Introduction to Social Network Methods. Riverside, USA: University of California.
- Hepburn, S., & Simon, R. (2010). Hidden in plain sight: Human trafficking in the United States, Gender Issues 27 (1-2) 1–26.
- Hu, X., and Liu, H. (2012). Text analytics in social media in Mining Text Data, Aggarwal, C., & Zhai, C. Eds: Springer, pp. 385-414.
- Human Rights First. (2017). Fact Sheet: Human Trafficking by the Numbers
- Hughes, M. (2001). The impact of the use of new communications and information technologies on trafficking in human beings for sexual exploitation: A study of the users. Strasbourg: Council of Europe.
- i2 Group. (2010). Social Network Analysis Whitepaper: Analyst’s Notebook 8, issue 2.
- Ibanez, M., & Suthers, D. (2014). Detection of domestic human trafficking indicators and movement trends using content available on open internet sources, in: 47th Hawaii International Conference on System Sciences (HICSS), IEEE, pp. 1556–1565.
- ILO. (2017). Global estimates of modern slavery. Available online: https://www.ilo.org/wcmsp5/groups/public/—dgreports/—dcomm/documents/publication/wcms_575479.pdf. Accessed 28 December 2019
- Johnson, S., & Bowers, K. (2004). The stability of space-time clusters of burglary, British Journal of Criminology 44 (1) 55–65.
- Kapoor, R.; Kejriwal, M., & Szekely, P. (2017). Using contexts and constraints for improved geotagging of human trafficking webpages in Proceedings of the Fourth International ACM Workshop on Managing and Mining Enriched Geo-Spatial Data – GeoRich ’17. New York, New York, USA: ACM Press, 2017, pp. 1–6.
- Kejriwal, M., & Szekely, P. (2017). Information Extraction in Illicit Web Domains in Proceedings of the 26th International Conference on World Wide Web – WWW ’17. New York, New York, USA: ACM Press, 2017, pp. 997–1006.
- Kena, F., Srivatsav, K., Kevin, C., & Rashmi, K. (2014). Data analytics and human trafficking. M.C. Tremblay et al. (Eds.): DESRIST 2014, LNCS 8463, pp. 69–84. Springer International Publishing Switzerland.
- Kennedy, E. (2012). Predictive patterns of sex trafficking online, Thesis, Carnegie Mellon University.
- Knorr-Held, L., & Richardson, S. (2003). A hierarchical model for spacetime surveillance data on meningococcal disease incidence, Journal of the Royal Statistical Society: Series C (Applied Statistics) 52 (2) 169–183.
- Konrad, R. Trapp, A., Palmbach, T., & Blom, J. (2017). Overcoming human trafficking via operations research and analytics: Opportunities for methods, models, and applications. European Journal of Operational Research, 259(2), 733–745.
- Latonero, M., Berhane, G., Hernandez, A., Mohebi, T., & Movius, L. (2011). Human trafficking online: The role of social networking sites and online classifieds. Center on Communication Leadership & Policy. Research Series: 09/11.
- Marturana, F., & Tacconi, S. (2013). A machine learning-based triage methodology for automated categorization of digital media. Digital Investigation. 10, 193-204.
- Merdian, H.; Perkin, D.; Webster, S.; McCashin, D. (2019). Transnational child sexual abuse: outcomes from a roundtable discussion. International Journal of Environmental Research and Public Health. 16, 243.
- Philip, A., Zheming, Z., Longzhi, Y., & Yanpeng, Y. (2019) An Intelligent Online Grooming Detection System Using AI Technologies. In: 2019 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), New Orleans
- Sarkar, S. (2015). Use of technology in human trafficking networks and sexual exploitation: A cross-sectional multi-country study. Transnational Social Review. Vol. 5, No.1.
- Silva, D., Philpot, A., Sundararajan, A., Bryan, N., & Hovy, E. (2014). Data integration from open internet sources and network detection to combat underage sex trafficking in Proceedings of the 15th Annual International Conference on Digital Government Research – dg.o ’14. New York, New York, USA: ACM Press, 2014, pp. 86–90.
- Szekely, P., Knoblock, C., Slepicka, J., et al. (2015). Building and using a knowledge graph to combat human trafficking. M. Arenas et al. (Eds.): ISWC 2015, Part II, LNCS 9367, pp. 205–221. Springer International Publishing Switzerland.
- UNODC. (2015). Study on the effects of new information technologies on the abuse and exploitation of children. Available online: https://www.unodc.org/documents/Cybercrime/Study_on_the_Effects.pdf. Accessed 29 December 2019.
- The Internet Watch Foundation. (2019). IWF Annual Report 2018: Once Upon a Year. April 2019. Available online: https://www.iwf.org.uk/sites/default/files/reports/2019-04/Once%20upon%20a%20year%20-%20IWF%20Annual%20Report%202018.pdf Accessed 29 December 2019.
- Territo, L., & Kirkham, G. (2009). International sex trafficking of women & children: Under- standing the global epidemic, Looseleaf Law Publications, 2009.
- Wang, H., Cai, C., Philpot, A., Latonero, M., Hovy, E., & Metzler, D. (2012). Data integration from open internet sources to combat sex trafficking of minors in Proceedings of the 13th Annual International Conference on Digital Government Research ’12. New York, New York, USA: ACM Press, 2012, p. 246.