Big data

Big data is generally defined as enormous data sets that cannot ‘be perceived, acquired, managed, and processed by traditional information technology and software/hardware tools within a tolerable time’ (Chen et al., 2014, p. 173). In other words, big data is defined by the voluminous size of the rows and columns of the data within the file, its increasing scale, and its inability to be handled by conventional technologies (Williams et al., 2017; Wu, Zhu, Wu, & Ding. 2014). This description is not uniform across all stakeholders, as different actors and disciplines hold varying conceptualisations of big data based on the particularaspects they value the most (Chen et al., 2014; Smith et al., 2017).

Numerous industry sectors and fields have shown interest in the potential utility of big data. For instance, Google has been reported to processes data requests equalling hundreds of petabytes (PB) every month, while Facebook generates log data of over 10PB a month. Other companies such as Baidu, a Chinese internetsendees company, have reported processing data at tens of PB per day (Chen et al., 2014). A 2011 report from the International Data Corporation (EDC) indicated that the total amount of data created and copied in the world was 1.8 zettabyte (ZB), an increase in volume of nearly nine times since 2006 (Chen et al., 2014).

If such quantities of data can be properly harnessed for analysis, substantial improvements on the productivity and competitiveness of both private and public organisations can be had, while also generating huge benefits for its consumers (Chen et al., 2014). For example, if big data could be used to its full potential by the US medical industry, it could potentially net over USD 300 billion and redttce expenditures of the US healthcare system by over eight percent (Chen et al., 2014). Private retailers that could fully utilise big data may also be able to increase then- profit by more than 60 percent (Chen et al., 2014).

In order to maximise the power and potential of big data, a fundamental change in the computing architecture and large-scale data processing mechanism is needed (Chen et al., 2014). To this effect, IBM has invested over USD 16 billion on more than 30 acquisitions related to big data since 2005 (Chen et al., 2014). These acquisitions consist of procuring other companies specialising in analyticsbased data processing. One such example is IBM’s acquisition of Aspera in 2014 - a company specialised in securely moving massive data files around the world using fast transmission speeds. Other examples include acquiring Vivisuno (a leading provider of big data discovery and navigation) in 2012 and Cleversafe (a leading developer in big data storage) in 2015. Similarly, the Obama administration had announced a USD 200 million investment to launch the ‘Big Data Research and Development Plan’ in March 2012, making it the second major-scientific and technological development initiative since 1993 (Chen et al., 2014).

While the advantages of big data are wide-ranging, its growth has produced many challenging obstacles that demand precise solutions. Specifically, it is imperative that analysts determine methods to ascertain values from data sets with enormous scales and rapid generation (Chen et al., 2014; Smith et al., 2017). This includes managing large quantities of unstructured data that require significant amounts of real-time-analysis. Additionally, data must be acquired from a wide pool of data sources in order to be combined for analysis (Chen et al., 2014; Williams et al., 2017). Finally, all of this information must be stored and retained with moderate requirements on hardware and software infrastructure.

Big data and traditional crime

Though many think of big data analytics as an issue for technology companies and industry, there is substantial value in operationalising big data and data mining analyses to the criminal justice system to combat modem crime problems (Chan & Moses, 2017). Police departments worldwide have begun using big data, machine learning, and predictive analytics to both understand and prevent crime, giving them the opportunity to position police resources in response to anticipated dangers (Clark, 2017). For instance, when previous arrest records are amassed with real-time internet of things (IOT) data, such as censored cameras constructed to recognise gunshots, it becomes easier for law enforcement to diagnose problem locations and understand the conditions in which crimes flourish (Clark, 2017). By applying predictive analytics and machine learning to large data sets, police departments are able to improve then ability to forecast both where and when violent crimes will emerge and ensure that they have the adequate resources in place to prevent such behaviours from occurring (Clark. 2017).

An excellent example of such a strategy is IBM’s HuncliLab, which is a geographic prediction tool that uses data modelling to identify various risks and dangers across specific areas (Clark, 2017). At-risk areas are marked on-screen, while suggestions for action are displayed along the side, such as dispatching a high visibility police patrol car to take control of the situation and deter criminals. This information is aggregated into a ‘decision support system’, made available to police officers in the field (Clark, 2017). IBM reports that this predictive approach has worked best against robberies, theft, and burglaries, as in Manchester, England, where its implementation was linked to reported reductions of robberies, burglaries, and theft from vehicles (Clark. 2017).

While tools such as HuncliLab are effective on certain types of crime, it is worth noting that other crimes may be much harder to prognosticate. This could be due to the random and isolated nature of various incidents; or because then-offenders operate clandestinely with the help and protection of larger organisations (Clark, 2017). In such scenarios, big data can still be used to predict the risks of crime. For example, digital traces from various online communication tools (e.g., social media, emails, webpages visited) can be analysed to predict potential attacks before they happen. A major challenge to this approach, however, is the volume of content. Not only can filtering (e.g., finding alarming content, trends.

Using big data to understand cybercrime 89 or patterns) through such data be time consuming, but the analyses can be misleading, as various online content may contain exaggerations and/or inaccuracies in meaning. Identifying whether certain items are truly predictors of crime can thus be difficult without deeper insight (Clark, 2017).