Abstract Purpose Previous research has documented changes in crime trends during the COVID-19 pandemic. However, less is known about crime trends after the pandemic and among minors. In this study, we investigate changes in crime trends among Italian minors during and after the COVID-19 pandemic. Methods The data for the current research consist of reports of crimes involving minors, submitted to judicial authorities by Italian law enforcement agencies, covering the period from 2004 to 2023. Using simple moving averages, observed crime rates were compared to expected values. Results Overall, we found that the number of reports of most types of crime did not change significantly during or after the COVID-19 pandemic. During the pandemic period, only child sexual abuse, theft, and drug trafficking and possession significantly decreased and, in the post-pandemic period, quickly returned to pre-pandemic baseline values. In contrast, the post-pandemic period was marked by an increase in attempted murder, physical assault, deliberate injury, threats, and robbery. Conclusions We conclude that the COVID-19 pandemic may have had both criminogenic and anti-criminogenic effects. Moreover, the increase in certain types of crimes in the post-pandemic period may be attributed to marginalization, socioeconomic disparities, and economic hardships.

A comparison of the racial composition of police stops to the entire population of a city or jurisdiction is frequently cited as evidence of racial bias in proactive policework. This article argues that using base population is naïve to the realities of the distribution of crime and policing. Using the example of Philadelphia, PA (USA), the impact of different benchmarks to estimate racial disparity in stop data is demonstrated. The range of alterative benchmarks include the spatial distribution of calls for service, the locations of violent crimes, and the demographic composition of suspects in crime as reported by the public. The article concludes by arguing that if cities ask police departments to prioritize certain problems and places, benchmarks to which police are held accountable should better reflect those priorities.

PurposeOffenders often select partners based on shared characteristics such as age, sex, or ethnicity, a phenomenon known as homophily. At the same time, co-offenders also face a challenge of choosing between trustworthy partners to maintain stable collaborations and useful partners who provide access to new skills and information. This study investigates how homophily shapes the structure of criminal networks and, consequently, the diffusion of information within these networks.MethodsUsing an Agent-Based Model, we simulate a population of offenders that select partners either randomly or based on high similarity preference. When two agents mutually select each other, they commit a co-offense, forming a social network and exchanging skills.ResultsCompared to the case of the random partner selection, the homophily-driven environment results in sparse networks with a higher number of repeated interactions between agents, but with a slower rate of skill exchange. Moreover, on the individual level, having many partners is more beneficial for diverse skill acquisition, but those partners should belong to different subgroups.ConclusionThe results provide insights into how offender preferences shape the structure and dynamics of criminal networks, particularly in relation to opportunities for collaboration and skill acquisition. The findings highlight a key trade-off introduced by homophily. Although it promotes stable partnerships, it restricts the exchange of information across the broader network.

Every year, hundreds of case files from police investigations and court cases are produced. The challenge of manual processing and analysing the vast volume of documentation resulting from investigations have prompted law enforcement agencies to embrace Natural Language Processing (NLP). Despite this, a systematic approach to automatically retrieve relevant data from police case files is lacking. To address this gap, we propose a methodology to process police case files and perform criminals’ role classification, criminals’ ties classification, and value network inferring through syntactic pattern recognition and BERTopic. The results reached allow us to affirm that the inferred value networks can congruently depict the roles and the connections between them. Although further research is needed, this methodology can propel the automation of the extraction of criminal networks’ dynamics implemented by law enforcement agencies, allowing them to refine and improve the strategies employed to disrupt criminal networks.

BackgroundFraud is a prevalent offence that extends beyond financial loss, impacting victims emotionally, psychologically, and physically. Advances in online communication technologies continue to create new opportunities for fraud, and fraudsters increasingly using these channels for deception. With the progression of technologies like Generative Artificial Intelligence (GenAI), there is a growing concern that fraud will increase in scale using these advanced methods, with offenders employing deep-fakes in phishing campaigns, for example. However, the application of AI, particularly Natural Language Processing (NLP), to detect and analyse patterns of online fraud remains understudied. This review addresses this gap by investigating the potential role of AI in analysing online fraud using text data.MethodsWe conducted a Systematic Literature Review (SLR) to investigate the application of AI and Natural Language Processing (NLP) techniques for online fraud detection. The review adhered to the PRISMA-ScR protocol, with eligibility criteria including language, publication type, relevance to online fraud, use of text data, and AI methodologies. Out of 2457 academic records screened, 350 met our eligibility criteria, and 223 were analysed and included herein. ResultsWe discuss the state-of-the-art AI and NLP techniques used to analyse various online fraud categories; the data sources used for training the AI and NLP models; the AI and NLP algorithms and models built; and the performance metrics employed for model evaluation. We find that the current state of research on online fraud is broken into the various scam activities that take place, and more specifically, we identify 16 different frauds that researchers focus on. Finally, we present the most recent and best-performing AI methods employed for detecting online scams and fraud activities. ConclusionsThis SLR enhances academic understanding of AI-based detection methods for online fraud and offers insights for policymakers, law enforcement, and businesses on safeguarding against such activities. We conclude that existing approaches focusing on specific scams are unlikely to generalise effectively, as they will require new models to be developed for each fraud type. Furthermore, we conclude that the evolving nature of scams limits the effectiveness of models trained on outdated data. We also identify that researchers often omit discussions of the limitations of their data or training biases. Finally, we find issues in the consistency with which the performance of models is reported, with some studies selectively presenting metrics, leading to potential biases in model evaluation.

BackgroundThis study examines the impact of the COVID- 19 pandemic on policing, focusing on changes in calls for service and spatial and demographic patterns of demand, and the experiences of call handlers. It explores how policing and community behaviours are adapted under crisis conditions. By examining shifts in demand and police response during the pandemic, we offer insights into how policing strategies and community behaviours evolved.MethodsThe study employs a mixed-methods approach, combining quantitative analysis of call data with qualitative interviews. The dataset covers calls for service from 2015 to 2020, aggregated at neighbourhood level. We used time series forecasting to create a counterfactual against which to compare observed data. Spatial analysis was performed using a Gini coefficient and Location Quotient to measure concentration within LSOAs and by linking call data with the Index of Multiple Deprivation to consider socio-demographic shifts. Thematic analysis of semi-structured interviews with call handlers examined their experiences.ResultsCall volumes dropped overall, but reports of anti-social behaviour (ASB) related to lockdown breaches and drug incidents increased, and became less spatially concentrated. Call handlers reported stress related to managing complex public health-related queries but a positive ability to resolve more calls remotely. They also reported changes in the nature of calls, such as around domestic incidents and harassment. Police attended a higher percentage of calls and reduced time spent on scene.ConclusionsCall patterns during the pandemic shifted in nature, spatial distribution, and socio-demographic trends, highlighting the need for ongoing monitoring and adaptive resource allocation. Insights from call handlers are crucial for understanding these changes and guiding strategies to support staff and respond effectively to evolving community needs.

Background settingDomestic violence (DV) is a major public health problem and a violation of human rights. To date, research on DV has predominantly focused on women as victims and men as perpetrators. Male DV victims particularly in non-intimate relationships have received little attention in the literature. This study represents the first attempt to report on DV among male victims in non-intimate relationships using population-level data.MethodsThis is a population-level retrospective observational study using data extracted from a large sample of police-attended narratives in New South Wales (NSW) from 2005 to 2016 using rule-based text mining.ResultsFrom 18,611 DV events involving non-intimate relationships, most of the Persons of Interest (POIs)—individuals suspected or charged with a DV offence—were male (78%) and members of the victims’ family (26.8%, cousins, uncles and aunts). A total of 42 different types of abuse were identified in 74.3% (n = 13,832) events, the most prevalent being physical abuse with assault (unspecified) accounting for half of the cases (53.9%, n = 7462) and punching for more than one third of cases (35.4%). Almost half of DV events (46.3%, n = 8616) recorded injury type to the victim, the most common being cut(s) (43.6%, n = 3754), followed by swelling (19.9%, n = 1716), and bruising (19.5%, n = 1679). A total of 2,903 (15.6%) events had a mental illness mentions for the POIs and 857 (4.6%) for the victims, with 23 different mental illnesses recorded. Schizophrenia and dementia were the most common mental illnesses among POIs (13.6%) and victims (13.0%), respectively.ConclusionsThis study provides new insights and empirical evidence on abuse types, perpetrator-victim relationships, victim injuries and mental illness on DV events involving adult male victims in non-intimate relationships. The findings form an important evidence base to trigger further research in the future.

ObjectiveThis study examines Bitcoin generator (BG) websites on the clear and dark web. It focuses on their prevalence, revenue, and associated warnings, as these sites are suspected scams.MethodData for the study was gathered from the Dark Web Monitor and Iknaio Cryptoasset Analytics. A four-step process was used to identify BG sites and their Bitcoin addresses from 2 million dark websites.ResultsWe found 832 dark web BG sites. The monetary revenue from a dark web BG site is approximately 1/3 smaller per Bitcoin address than from a clear web BG site. There is a concentration of revenue at a few BG sites. Only 24% of Bitcoin addresses on dark web BG sites have ever had money deposited on them. On the dark web, the top three clusters of crypto addresses account for 35% of the total revenue. On the clear web, the top three clusters account for 52% of the total revenue. The longer BG sites are online, the higher the revenue. There are hardly any warnings against BG sites.ConclusionOur results fit the Rational Choice model of crime: the revenue is modest, but the effort of the offenders is also limited.