C3-Sex: A Conversational Agent to Detect Online Sex Offenders

Abstract

Prevention of cybercrime is one of the missions of Law Enforcement Agencies (LEA) aiming to protect and guarantee sovereignty in the cyberspace. In this regard, online sex crimes are among the principal ones to prevent, especially those where a child is abused. The paper at hand proposes C3-Sex, a smart chatbot that uses Natural Language Processing (NLP) to interact with suspects in order to profile their interest regarding online child sexual abuse. This solution is based on our Artificial Conversational Entity (ACE) that connects to different online chat services to start a conversation. The ACE is designed using generative and rule-based models in charge of generating the posts and replies that constitute the conversation from the chatbot side. The proposed solution also includes a module to analyze the conversations performed by the chatbot and calculate a set of 25 features that describes the suspect’s behavior. After 50 days of experiments, the chatbot generated a dataset with 7199 profiling vectors with the features associated to each suspect. Afterward, we applied an unsupervised method to describe the results that differentiate three groups, which we categorize as indifferent, interested, and pervert. Exhaustive analysis is conducted to validate the applicability and advantages of our solution.

Cybercrime: A systematic review of the literature

“…In the social context (table 3), the presence of digital tools as a means of committing crimes is evident: scripted URLs (7 references), malware (12 references), cyberspace (5 references), online games (3 references), social networks (5 references), phishing (4 references), online dating (2 references) and adult pages (4 references): [18] Malware Shapira, 2021 [16] Maskun, 2020 [44] Sviatun, 2021 [18] Phishing Cascavilla, 2021 [29] Alzubaidi, 2021 [28] Maskun, 2020 [44] Cascavilla, 2021 [29] Zahrah, 2020 [50] Gonzales, 2020 [55] Onlin e dating Shapira, 2021 [16] Basuchoudhary, 2019 [58] Adult pages Onuoha, 2020 [22] Cyberspace Seung-Yeop, 2021 [3] Couzigou, 2019 [61] Onlin e games Sviatun, 2021 [18] Table 7 shows the tools used for the commission of economic crimes, being malware with 15 references and phishing with 9 references the most preponderant: Ibañez, 2020 [43] o…”

“…Cybercrime: [18], [30], [31], [33], [34], [35], [38], [44], [72]; Child Pornography: [32], [36], [41], [43], [45], [57], [60]; Harassment: [46], [64]; Sexual violence: [54], [59]; Extortion: [5], [55], [56].…”

Child Safety and Protection in the Online Gaming Ecosystem

“…Nonetheless, most research [29]- [52] has not been conducted in the context of online gaming but rather on chat logs related to social media websites, instant messaging applications, and public chat rooms. The distributed platform chatbots in [29], [30], [36], [42] utilize natural language processing (NLP) and artificial intelligence markup language (AIML) to process the conversation for text content and emotion classification in addition to opinion classifiers to detect predatory behav-ior on chatting platforms. The most widely used technique for detecting child predators on chatting platforms is text classification, which utilizes the two available data sources reported in [53], [54].…”

“…The most widely used technique for detecting child predators on chatting platforms is text classification, which utilizes the two available data sources reported in [53], [54]. ML methods used to detect cyberbullying [27], [28], [31], [48] and sexual predatory behavior [25], [29], [30], [32]- [40], [42]- [47], [49]- [52] remain the main focus of research over the last 10 years. In [55], a brief survey of ML algorithms was used to detect child grooming behavior on social media between 2007 and 2016.…”

BOTTER: A framework to analyze social bots in Twitter

“…Additionally, strong lines are making advances in the fight against social bots [10], those semi or fully automated profiles that programmatically launch interactions on Online Social Networks (OSN) [11]. Although there are chat, informative or joke social bots [12], the majority of studies tend to spot malicious fake accounts that create large number of iterations and alter the natural dynamics of social media [13]. In this sense, academic evidences mainly focus on the detection of these illegitimate accounts, performing high-level statistical and temporal analysis to compare with legitimate users [14].…”