Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
THE RISE OF TECHNOLOGY IN CRIME PREVENTION:
OPPORTUNITIES, CHALLENGES AND PRACTITIONERS’
PERSPECTIVES
A PREPRINT
Dario Ortega Anderez
School of Science and Technology
Nottingham Trent University
Nottingham, UK, NG11 8NS
dario.ortegaanderez02@ntu.ac.uk
Eiman Kanjo
School of Science and Technology
Nottingham Trent University
Nottingham, UK, NG11 8NS
eiman.kanjo@ntu.ac.uk
Amna Anwar
School of Science and Technology
Nottingham Trent University
Nottingham, UK, NG11 8NS
amna.anwar2016@my.ntu.ac.uk
Shane Johnson
Dept of Security and Crime Science
UCL Jill Dando Institute of Security and Crime Science
London, UK, WC1H 9EZ
shane.johnson@ucl.co.uk
David Lucy
Metropolitan Police
London, UK, SW1A 2JL
David.M.Lucy@met.police.uk>
February 9, 2021
ABS TRAC T
Criminal activity is a prevalent issue in contemporary culture and society, with most nations facing
unacceptable levels of crime. Technological innovation has been one of the main driving forces
leading to the continuous improvement of crime control and crime prevention strategies (e.g. GPS
tracking and tagging, video surveillance, etc.). Given this, it is a moral obligation for the research
community to consider how the contemporary technological developments (i.e. Internet of Things
(IoT), Machine Learning, Edge Computing) might help reduce crime worldwide. In line with this,
this paper provides a discussion of how a sample of contemporary hardware and software-based
technologies might help further reduce criminal actions. After a thorough analysis of a wide array of
technologies and a number of workshops with organisations of interest, we believe that the adoption
of novel technologies by vulnerable individuals, victim support organisations and law enforcement
can help reduce the occurrence of criminal activity.
Keywords Crime prevention ·Emerging technologies ·Pervasive computing
1 Introduction
Criminal activity continues to be a major concern in modern civilisations, with most nations facing unacceptable levels
of crime and delinquency [
1
]. Crime is a general term which embodies a wide array of criminal actions. Across
offence types, academic literature and official statistics indicate significant variation in both risk and reporting rates. For
instance, statistics from the UK Office for National Statistics [
2
] suggest that vehicle-related thefts are amongst the most
frequently experienced crimes, whereas homicides are relatively rare. With regards to reporting rates, while these are
high for crimes such as domestic burglary, domestic abuse (DA) has been identified as one of the most under-reported
arXiv:2102.04204v1 [cs.CY] 26 Jan 2021
APREPRINT - FE BRUARY 9, 2021
Figure 1: Paper outline with the main technologies discussed.
offence types. A common assumption is that victims of DA, especially women who are abused by their partners, are
reluctant to report domestic violence (DV) to the authorities [3, 4, 5].
While crime in its multiple forms has generally declined over the last two decades [
6
], there are notable exceptions. For
instance, as per UK 2019 figures, violence against the person has shown a relevant increase as compared to the previous
year [
2
]. Increases in specific types of crime are often attributable to changes in opportunity. For instance, increases in
online activity have led to significant opportunities for cyber-enabled (made easier by the internet, such as fraud) or
cyber-dependent (only possible because of the internet, such ransomware attacks) crime. For other criminal offences,
such as domestic abuse, while figures may be increasing, part of the reason for the trends observed in the UK Police
data is that the gap between what is experienced and what is reported is closing [
7
]. That is, improvements on police
recording practices and the support services provided appear to be encouraging victims to come forward more than they
once did. This is critical as crime cannot be prevented without a clear understanding of the scale of a specific offence
and of how these are committed. Put differently, the challenges associated with addressing crime are not limited to
reducing offending. Providing (and signposting) support services for those affected, and capturing the true scale of the
problem can play a vital role in preventing further criminal activity.
As alluded to above, technological innovation can generate opportunities for crime. However, technology has also been
one of the driving forces that has enabled the improvement of approaches to crime prevention, and poling more generally
[
8
]. For example, the first police technology revolution, which incorporated the use of two-way radio, automobiles
and telephones, facilitated substantial change in the way police organisations operate [
9
]. As several authors suggest
[
8
,
10
,
9
], it is now time for the second revolution, with such a revolution implicating once again a change in police
administration and organisation. The use of Information Communication Technology (ICT) has and is likely to continue
to be capital in terms of understanding crime problems and responding to them. However, to date, compared to industry,
in policing the exploitation of developing technologies has been somewhat limited in the crime reduction enterprise.
This paper provides a review of the different technology-driven solutions that have been employed for the purposes of
crime prevention in the last few decades, as well as a critical analysis of how contemporary hardware and software-based
technologies, such as ubiquitous sensors and machine learning, could be used for improving the reporting and the
prevention of criminal activities in the (near) future. In what follows, the technical advantages and disadvantages of a
wide array of sensing devices are discussed in the context of crime prevention (see Fig. 1), with the aim of identifying
new research directions and encouraging research activity in the field.
In addition to the analysis of opportunities new technologies present to monitor and prevent crimes, a number of
sandpits and co-design workshops with victim support, policing and end users organisations were conducted. These
discussions have also helped in understanding the relationship between the different crime stakeholders (see Fig. 2), as
well as how technology can not only support victims but also prevent or reduce crime levels by monitoring offenders
behaviour and their interaction with the surrounding environment.
The remainder of this paper is organised as follows (see also Fig. 1). Section 2 reviews current applications of pervasive
computing technologies used in crime prevention. Section 3 discusses the challenges and opportunities identified for
the employment of pervasive technologies to prevent, detect and report criminal activity. Section 4 presents the main
challenges for the adoption of novel technology-driven crime prevention tools. Section 5 concludes the manuscript with
a summary of the main findings and provides recommendations for future work on the developments of technological
tools for crime prevention.
2
APREPRINT - FE BRUARY 9, 2021
Figure 2: The relationship between crime stakeholder and the environment.
2 Technology-driven Solutions for Crime Prevention
The advent of pervasive computing and the IoT provides new potential opportunities for technology-driven solutions
for crime prevention [
11
,
12
]. This section discusses the working principle of such technologies as well as (where
appropriate) how these are currently used for preventing crime. The section is organised according to the main source
of information or sensing device used by each of the discussed technologies.
2.1 Camera-based Technologies
In recent years, Closed Circuit Television (CCTV) surveillance has emerged globally as a mainstream crime prevention
measure [
13
]. CCTV technology is already present in many public places such as railway stations, airports, office
buildings and on the street. Moreover, as early as 2005, estimates suggest that in the United Kingdom there were already
over 4 million cameras in place – an approximate ratio of 1 camera per 14 citizens [14].
While CCTV cameras were formerly employed as a means to report crime, support police officers with prosecution
processes or as a supporting evidence in court of justice, the increasing developments in computer vision and machine
learning provide opportunities for investigating CCTV surveillance as a crime prevention mechanism [13].
CCTV cameras are typically intended to reduce crime through several mechanisms. The first is to deter offenders
who fear being identified committing a crime on camera. The second is to provide law enforcement agencies with
an opportunity to identify crimes in progress or those that are about to happen, to provide them with an opportunity
to intervene. Where these two mechanisms fail, CCTV cameras can record evidence that a crime occurred, who was
involved and what took place, which can be used in criminal prosecutions. The first mechanism relies on offenders
perceiving that CCTV is effective, otherwise there will be no perceived risk. The second relies on staff monitoring
(possibly many) camera feeds and being able to identify suspicious activity. The third relies on the camera(s) being able
to capture an event in progress, which may require it to pan and tilt accordingly. It is not hard to imagine why failures
associated with human error might emerge, given the difficulties associated with these tasks. However, developments in
computer vision and machine learning, which are leading to a continuous improvement in applications such as object
and face recognition [
15
,
16
] or gait analysis[
17
,
18
], clearly provide opportunities for CCTV surveillance that might
enhance its effectiveness [13].
CCTV surveillance systems are constantly being upgraded to incorporate the latest soft technology features [
12
]. These
include on-the-edge feature extraction and machine learning classification algorithms. As a result, it is now possible to
extract different patterns and personal bio-metrics from video sequences, which can be a posteriori used for person
identification. In this sense, CCTV technology has been investigated in different applications (see Fig. 3). These
include the automatic detection of suspicious anomalies such us unattended bags in mass transit areas or crowded
venues [
19
], iris recognition-based security systems which deny access to buildings to unauthorised personnel [
20
],
intrusion detection systems (IDS) in unauthorised areas employing motion tracking techniques [
21
,
22
], automatic
robbery detection in banks via object and human posture detection [
23
], and even as a means of crowd detection and
congestion analysis for safety purposes [24].
3
APREPRINT - FE BRUARY 9, 2021
Figure 3: Applications of CCTV surveillance on crime prevention
2.2 Electronic Monitoring and Global Positioning System-based Technologies
Electronic Monitoring (EM) began in the early 1980’s in the US and spread rapidly after positive initial claims in
reducing control deficits in community supervision [
25
]. First EM systems employed radio frequency identification
(RFID) technology. RFID technology is based on a tag with a unique identifier which sends data to an electronic reader
through wireless radio frequency waves, enabling its identification and tracking. RFID technology was first employed
to confine offenders (or pre-trial defendants) to a particular location (usually their home) [
26
]. The work in [
27
] showed
that placement of sex offenders on EM programs reduced the likelihood of and postponed their return to prison. The
use of RFID technology was then extended to the protection of victims of domestic violence. Victims were provided
with receivers so that they were alerted when an offender was present within a pre-established control perimeter, which
was normally set to be around the victim’s home [28] (see also Fig. 4).
The second generation of EM technologies incorporated the use of Global Positioning System or GPS. GPS is a
satellite-based global navigation system which provides geo-location through the use of a network of satellites orbiting
the Earth at an altitude of approximately 20,200 km. To estimate the geo-location, a GPS receiver intercepts the signals
of at least three network satellites at regular intervals of time. A posteriori, based on the time it takes to receive each of
the satellite signals, the geo-location of the GPS receiver is calculated via trilateration. The use of GPS technology
as a crime prevention tool has gained increasing attention since late 1990’s [
29
]. The ability to customise exclusion
zones and provide instant alerts if these are violated, has extended the use of electronic monitoring to sex offenders and
post-work release offenders [
30
]. A pictorial example of how GPS technology is used in this context can be seen in
Fig. 4. Further applications, such as the tracking of terrorist suspects to gain insights into their spatial and temporal
behaviour, have recently been proposed [31].
Overall, as outlined by [32], EM is an effective tool towards preventing recidivism, specially for sex offenders.
Figure 4: GPS Geo-location .
4
APREPRINT - FE BRUARY 9, 2021
Figure 5: Social context monitoring via Bluetooth device scanning.
2.3 Short-Range Wireless Communication Technologies
Short-range wireless communication transceivers such as Bluetooth and WIFI are widely incorporated into many
portable and mobile devices, including laptops, mobile phones and smart-watches. During the manufacturing process, a
wireless module is assigned a unique identification (ID) in the form of a 48-bit Medium Access Control (MAC) address.
This address is then used to identify and authenticate a device when communicating with other wireless devices. For
example, Bluetooth (BT) devices can interact with other nearby BT devices within their signal range (
10m
to
100m
,
depending on the radio transceiver) by sending and receiving radio waves within a band of 79 different frequencies
centred at 2.45 GHz. Using such radio waves, along with the identification capabilities provided by the unique MAC
address assigned, a Bluetooth device can continuously monitor other Bluetooth devices nearby (within its signal range)
and also identify the type of device associated with such MAC address (i.e. whether it is a smart-phone or a laptop
for example). In addition, the Received Signal Strength Indicator (RSSI) provides an estimated measure of the power
present in a received radio signal. As shown by previous RF-based research [
33
,
34
,
35
], RSSI can then be used to
estimate the approximate distance a Bluetooth receiver is from a Bluetooth emitter.
Exploiting the above characteristics of this technology, various researchers have employed Bluetooth technology to
estimate the surrounding social context of a person [
36
] (see Fig. 5) or to estimate pedestrian flows at specific places
[
37
,
38
]. The authors in [
11
] discuss how such monitoring capabilities could be used to prevent or investigate the
kidnapping of infants and elementary school children. They note that kidnappings tend to take place when children
are alone, and that a system could be used to continuously log the Bluetooth devices near to a child and, when none
are detected, an accelerometer is used to monitor their activity. Ultimately, they suggest, a mobile application (where
the Bluetooth logs can be visualised) could be provided to the child’s parents so that they can monitor their children’s
social context throughout the day.
2.4 Audio-based Technologies
Audio-based technologies have been used for many years to anticipate criminal actions. Common examples include the
interception of calls by police officers (wire tapping) or the use of recording devices as evidence collection mechanisms
[
39
]. With the current advances in audio processing and machine learning techniques, whereby several parameters and
bio-metrics can be automatically computed from raw audio recordings, the scope of audio-based technologies have
experienced a great expansion. For instance, gun shot detection audio-based technologies have been recently evaluated
in the US [
40
]. Gun shot detection is a novel technology which employs a network of microphones, typically installed
in high crime areas, which discriminates gunshots from other types of noises and computes the spatial coordinates
of the location where the shot was fired. Besides, the presence of Intelligent Virtual Assistants (IVAs) or Chatbots
such as Amazon Alexa or Google Assistant is becoming increasingly popular [
41
]. Such IVAs incorporate speech
recognition capabilities allowing users for asking questions and making requests to different interfaces. In addition to
speech recognition, it is now possible to count the number of speakers in a conversation via speaker diarization [
42
],
to infer the sentiment (mood) of individuals by analysing their voice [
43
] or yet recognise a speaker by his/her voice
with considerably low error rates [
44
,
45
]. These advances clearly provide opportunities for implementing audio-based
crime prevention tools.
2.5 Affective Computing
Recent research has realised that there is a significant relationship between the physical health and emotional state
of an individual [
46
]. Given this, the field of affective computing is receiving increasing attention in the last few
years [
47
,
48
,
49
]. Affective computing or emotional intelligence is the study and development of systems for the
5
APREPRINT - FE BRUARY 9, 2021
recognition, processing and interpretation of human affects. Typically, this is performed with the use of wearable
devices or smart textiles by which various physiological signals related to stress levels are measured, processed and
interpreted. Electro-dermal activity (EDA) and heart rate variability (HRV) are two major examples of physiological
signals employed in affective computing. Although affective computing is still as its infancy, several studies have shown
that such signals can be translated into relevant features which ultimately lead to estimations of human stress levels. For
instance, the research studies in [
50
,
51
] have made use of the MIT Stress Recognition in Automobile Drivers Database
[52] to classify between three different stress levels (low, medium and high) in three different driving scenarios.
Although, to the best of our knowledge, affective computing has not been employed in the context of crime prevention,
the anticipation of high levels of stress could assist the prevention of crime. As suggested by literature in the field [
53
],
mental disorder and violence may each be rooted in the stress levels at which an individual lives. On another note,
mental disorder has been shown to increase the chances of committing crime. For instance, in the study conducted
by [
54
] -using a Swedish birth cohort selected at random- it was found that men with major mental disorders were
more than ten times more likely than men with no mental disorders to have criminal offense records and four times
more likely to be registered for a violent offense. Other work [
55
,
56
,
57
] suggest that, on average, victims of domestic
violence have poorer mental health, which can lead to further issues including depression or anxiety. Although as
aforementioned, sentiment analysis is still at its infancy, this technology clearly shows the potential to be explored as a
crime prevention mechanism.
2.6 Information-based Technologies
In addition to the technologies discussed above, a wide array of information-based crime prevention solutions have
been developed in the last years.
One of such technologies is Crime Mapping (CM), also known as hot-spot policing [
58
]. CM refers to the process of
conducting spatial analysis to map, visualise and analyse crime patterns[
59
]. This allows for the identification of crime
hot spots in conjunction with other crime trends and patterns[
60
,
61
]. Such information can then be used to optimise
the location of human or/and technological resources. As various works suggest, the identification of hot spots is an
effective software-based technology for optimising the use of resources and ultimately prevent crime [62, 63].
Risk assessment is another key information-based technology for crime prevention. Risk assessment is used to
assess the risk of recommitting crime by offenders under correctional control. According to the survey conducted in
[
64
], a majority of serious crimes are committed by a small fraction of people during the first months of probation
parole. Risk assessment tools make use of predictive models to identify such subgroup of people so that appropriate
surveillance/supervision is granted to those cases. Likewise, information technology is used to identify the likelihood of
a terrorist attack or a serious violent event occurring at certain places, including schools, airports or train stations among
others [
12
]. Another application where information technology has been adopted to prevent crime is in the development
of computer software to track individuals’ interactions on various social media sites [
65
]. The monitoring of such
suspect’s interactions is then used to identify abnormal behaviours which can potentially be related to crime intentions.
2.7 Crime Prevention Mobile Apps
The number of downloads of mobile apps has been steadily growing worldwide for the past decade. Along with this,
the average time per day spent by adults on their smartphones is growing at a fast pace and this is forecasted to continue.
For instance, in the UK, that figure is expected to go over 4 hours in 2021 [
66
]. With this in mind, the development
of mobile apps as a means of crime prevention seems highly beneficial for promoting and enhancing social safety. In
this context, a wide array of applications have been proposed to help either with the prevention of crime or with the
reporting of crime that has already occurred. Apps for crime prevention can be broadly divided into two categories,
namely apps to be adopted by policing institutions and those directed to the general public. The remainder of this
section presents a wide array of mobile apps for crime prevention with a special focus on those related to the reporting
of emergency situations.
Mobile Apps for Emergency Situations:
Numerous mobile apps have been developed in recent years to facilitate the
communication of emergency and panic situations to close relatives and policing institutions. For instance, Circle of 6
[
67
] is a mobile app that enables its users to quickly contact a user-defined list of six people and share their current
location along with an emergency message. In a similar way, Noonlight [
68
] provides a mechanism to alert emergency
institutions by pushing and holding a button embedded in the app. When the button is released, the user is asked
for her/his pin number. A posteriori, the user just needs to enter the pin to indicate he or she is safe. If the pin is
not provided, the situation is immediately reported to emergency institutions along with the live GPS geo-location of
the user. A similar working principle to that of Circle of 6 and Noonlight is provided by Silent Beacon [
69
]. In this
case, a portable device in the form of a panic button is provided alongside the app. Once the button is triggered, the
6
APREPRINT - FE BRUARY 9, 2021
Figure 6: Summary of indicative popular crime prevention apps in the Google Play store
device communicates the action to the smartphone app which then automatically calls 911 while communicating the
emergency situation and the location to a user-defined list of contacts. A similar functionality is offered by Red Panic
Button [
70
], which in addition to the above features, incorporates the automatic publication of emergency messages on
Twitter. Another example is bSafe [
71
], which allows a list of user-selected contacts to track the user’s way home using
GPS information. In addition, as with the apps presented above, it allows the communication of emergency situations
while also collecting evidence in the form of audio and video recordings.
Fig. 6 lists all the apps reviewed in this section including their app icon, number of downloads and user ratings.
Mobile Apps (miscellaneous):
In addition to the apps to aid in the reporting of emergency situations, a number of
further mobile apps which likewise aim at the prevention of crime in varying ways are available for the public. One of
such mobile apps is BrightSky [
72
] (10k+ downloads on GooglePlay, scores 4.0). BrightSky is a mobile app mainly
directed towards victims or potential victims of domestic abuse, being the first smartphone app to provide a UK-wide
directory of specialist domestic abuse support services. BrightSky enables users to locate their nearest support centre
by searching their area, postcode or current location. Other features include a tool to log incidents of abuse in text,
visual and audio forms. In addition, a number of questionnaires to assess the safety of a relationship as well as further
information and resources regarding domestic abuse, sexual consent, harassment and stalking are provided withing the
app. SpotCrime+ Crime Map [
73
] is a mobile app which provides a visual platform in the form of map with different
crime icons to keep the public aware of the crimes that occur in their area in real time by retrieving data from a US
crime database. With this information, the public can both avoid getting close to areas of high crime and provide any
evidence or statement concerning the crime of interest in their area. A similar approach is followed by the app Police
Crime Statistics Map [
74
], which provides a similar platform to that of SpotCrime but for UK citizens. In addition
to the above, various apps have also been developed to assist police officers in a wide array of their numerous work
duties. An example of these is Miranda Warnings Rights [
75
] which provides an easy to access reference to assist police
officers in mirandizing suspects.
7
APREPRINT - FE BRUARY 9, 2021
Figure 7: A number of modern edge devices with potential use in behaviour analysis for crime detection and prevention.
From the analysis of the above, it can be concluded that smartphone apps provide effective mechanisms to both victims
and police organisations to prevent crime in different ways. Further testing, development and promotion of crime
prevention apps can be crucial to reduce criminal activities under different circumstances.
2.8 Edge Computing and on Device Processing
Traditionally, the majority of the processing for data intensive applications was done on a central cloud to take advantage
of fast and powerful computing infrastructure. However, major issues concerning latency, security and privacy can be
identified in the use of cloud-based systems for crime prevention.
In this regard, a new trend of processing the data on the edge is emerging. The motivation behind edge computing
is that of performing the data processing as near as possible to the point of data production. With this, the privacy
and latency issues present in cloud-based systems can be significantly mitigated. AI is gradually finding its way into
embedding systems which are becoming smaller and less power demanding, while offering fast processing power and
low latency at an increasingly attractive cost.
A number of off-the-shelf edge computing devices suitable to carry out heavy signal processing and machine learning
applications are already available (see Fig. 7). For instance, both Nvidia and Google have recently released their
respective development boards, namely Jetson Nano and Google Edge TPU, with the aim of enabling users to develop
and run AI applications on the edge. In addition to their portability and the privacy advantages they offer, such
boards are supported by sophisticated development kits that consist of a SOM (System-on-Module) connected to a
development board which incorporate numerous connectors like USB and Ethernet to share the data gathered when
desired. Furthermore, the above devices also support major deep learning frameworks and tools such as TensorFlow.
The above, alongside current developments in the field [
76
], suggest that edge computing is finding its way into city
centers for the early detection and prevention of criminal actions.
3 Opportunities
This section discusses the potential use of the technologies presented in Section 2 for the prevention of criminal actions.
A summary of the information discussed in this section is provided in Table 1.
CCTV surveillance has been widely employed to prevent and report crime under different circumstances. The continuous
developments seen in pervasive computing as well as in computer vision, allow for the on-board computation of different
bio-metrics (i.e. face recognition or gait analysis) and for the extraction of patterns which are key to identify suspects
as well as activities taking place within its field of view. Despite the great potential shown by CCTV surveillance on
preventing various type of crimes in public and outdoor places [
13
], several drawbacks are found on its application
inside home environments (i.e. abusive relationships). First, after the installation, the device is typically visible. While
this is paramount for applications where the aim is to deter crime, where the aim is to collect evidence surreptitiously it
would be better to conceal the device. Second, a home CCTV surveillance system would require the installation of
numerous CCTV cameras, since the view field is limited to the room at where the camera is mounted. Third, CCTV
cameras are sensitive to lighting conditions and occlusion (i.e. aggressive behaviour in the night could not be detected).
8
APREPRINT - FE BRUARY 9, 2021
Table 1: A list of technology-driven solutions for the prevention of criminal actions.
System Technical Advantages Technical Disadvantages Potential Applications
CCTV-based Person identification Privacy concerns at home environments Collection of evidence
Activity Recognition Visibility Detection of unattended objects
Object recognition Lighting conditions Robbery prevention
Occlusion Physical violence prevention
Limited field of view Deterring offenders
High power consumption
Bluetooth-based Ubiquity Limited functionality on its own Social context analysis / Detection of clusters of people
Social context analysis MAC Randomisation Stalker detection
Low energy consumption RSSI Instability Home and vehicle robbery prevention
GPS-based Ubiquity Data unavailability Outdoors abuser tracking
NFC-based Ubiquity Activation by user Emergency alert mechanism
Reduced size Easy mechanism for accessing information
Audio-based Sentiment Analysis Lower degree of detail as compared to CCTV Evidence collection
Ubiquity Detection of violent actions (i.e. glass breaking)
Spherical sound field
Information-based Behavioural Analysis Long-term conclusions Abnormality detection in social media
User-dependent Hot-spot identification
Data-dependent Crime prediction
Ultimately, as outlined by numerous works, there is a common reluctance towards the adoption of video cameras in
home environments due to privacy concerns [
77
]. As a summary, video cameras alongside the latest machine learning
and pattern matching algorithms can be of great use to avoid crimes in outdoor settings as well as in private businesses
and public transport but alternative sensing technologies are preferred for home environments.
Short-range communication technologies such as WiFi and Bluetooth have been proven useful for analysing the social
context of a person [
36
,
11
]. For instance, by continuously logging the presence of Bluetooth devices around a device
and their respective Received Signal Strength Indicators (RSSIs), it is possible to know the alone and accompanied
periods of an individual. In addition, it is possible to infer the social context of a person by the analysis of the variation
in the number of surrounding devices across time. For instance, when a person is walking on the street, the social
context could be inferred by the changes on the number of devices detected across time. Likewise, when a person is
alone, the number of surrounding devices is not expected to vary. Furthermore, the closeness of an external device
to a person could be evaluated by the long-term analysis of the corresponding logged MAC address of such external
devices across time. This could be of great use to identify stalking scenarios, and more importantly, the periods of
time when the potential stalker and the potential victim are found alone. However, an increasing number of devices
make use of MAC randomisation. Therefore, to achieve such aim, hard settings to elude MAC randomisation should be
employed. This could be achieved by providing pre-programmed hardware such as wearable devices through which
a unique identifier is broadcasted. With this, a novel electronic monitoring system with a similar working principle
to that of the contact tracing utilised to prevent the spread of the COVID-19 could be deployed. In cases where
offenders are known to the police and have to use EM as a condition of their sentence, this system could be employed to
complement current systems which generally use GPS technology, providing altogether a more robust system which
would overcome the frequent lack of signal shown by GPS technology in indoor environments. Current portable devices
such as smart phones or smart watches already incorporate Bluetooth transceivers, allowing for the continuous log of
surrounding devices. In addition, current Bluetooth transceivers make use of a low energy protocol, namely Bluetooth
Low Energy (BLE), reducing the power consumption required to operate. A further potential application with the use of
Bluetooth device scanning is the prevention of robberies at home environments. To do so, Bluetooth Beacons could
be installed at home environments to detect surrounding devices which remain within the Bluetooth antenna range
for suspiciously long periods of time or/and at a suspicious time, raising an alarm when appropriate. Furthermore,
short-range communication technologies could be employed for the monitoring of different areas through the detection
of intrusive scenarios.
GPS technology can be of great use to ensure security distances are met. Reported cases of DV in which the offender
is forced to wear a GPS tracking device are a good example. However, despite the fact that most portable devices
incorporate GPS technology, personal geo-locations are only shared with services the users opts to share the data with,
therefore being not publicly available. Given this, GPS technology can be considered of little use for unreported cases
9
APREPRINT - FE BRUARY 9, 2021
of crime. On the contrary, short-range communication technologies constantly broadcast a signal to the surrounding
devices, thus overcoming the drawback shown by GPS technology in this context.
Short-range tagging technologies such as Near-field communication (NFC) technology are gaining increasing attention
and made available in most of the latest generations of smart phones. NFC technology is a set of communication
protocols for short-distance communications (4cm or nearer) between two electronic devices, namely a NFC reader
and a NFC tag. Such communication is achieved through the NFC reader sending a signal to a pre-programmed
battery-free NFC tag which sends back its function or application to the reader. Common examples of the use of NFC
technology include contact-less payment and door card readers. Given the reduced size of NFC tags, these could be
potentially embedded in a wide range of everyday objects such as clothes, key rings or furniture to be used as an alerting
mechanism in cases of emergency. Likewise, tag tapping could be used to trigger evidence collection mechanisms
such as microphones or video cameras in a quick, easy and efficient way. NFC technology would be therefore a great
complement to the current emergency reporting and evidence collection mobile apps presented in Section 2.7, which
will enable the triggering of the different apps’ built-in actions on the quiet.
Audio-based technologies have been employed for many years as a crime prevention tool through the use of wire
tapping and recording devices [
39
]. In addition, novel applications such as gunshots [
78
] and screaming/shouting
detection [
79
] are surging with the advent of machine learning and pervasive computing. From the psychological
perspective, as compared to CCTV surveillance, audio monitoring is considered as a less invasive way of monitoring
individuals [
80
]. This therefore indicates the privacy concerns found in the use of video surveillance systems in home
environments are mitigated by the use of audio sensing devices, making them an attractive mechanism for the collection
of evidence. In addition, as discussed in Section 2.4, the current advances in audio processing and machine learning
allow for the computation of various elements which can be crucial to prevent domestic violence. Let’s consider a
violent argument between a potential victim of domestic abuse and her/his abuser. Through audio diarization, their
voices can be separated, therefore allowing for the analysis of each of the audio streams separately. Bad language within
the conversation could be identified through the use of speech recognition. In addition, the mood (i.e. anxious, violent,
sad) of each person could be estimated through the use of audio sentiment analysis techniques. Overall, audio-based
technologies can be of great use as a means of evidence collection. Further advances in audio and natural language
processing would grant this technology with crime prevention capabilities by the identification of emergency situations
inferred from sound.
Information-based crime prevention systems make use of various data sources to predict the risk of a crime being
committed. For instance, as previous work shows [81], there is a common tendency from abusers to isolate victims of
domestic abuse by controlling and monitoring their electronic devices and social media, with several reported cases
of account hijacking, or even destroyed devices. Given this, information-based technology could be employed to
identify deviations from the regular personal social media behavioural pattern from a potential victim of domestic
abuse, therefore providing clues for the detection of further criminal actions. Likewise, analysing the social media
behaviour of individuals, it is possible to employ probabilistic models to estimate the probability for a specific person to
commit crime. In addition, crime mapping techniques [
82
] which allow the identification of crime hot spots could be
employed to optimise the use of personal and technological resources. Although the information gained from the above
techniques could benefit the identification of potential criminal actions, various important limitations are found. First,
the use of trend analysis techniques may require long-term studies. That is, the behaviour of individuals may need to be
monitored for some time to establish a useful profile of activity. This can imply a long delay between the occurrence
of the criminal actions and the conclusions drawn from the analysis. Second, in the context of personal social media
behavioral analysis, a previous active use of social media by subjects of interest may be required. Third, may people
seek to remain anonymous online by using pseudonyms [
83
]. Where they do, it will be difficult to develop profiles of
their activity that could be used for the purposes of crime prevention. To use such data, it may be necessary to have
an individual consent. Fourth, even if deviations from personal patterns are identified, care would need to be taken in
interpreting these as it cannot be assumed these are surely caused by the experience of abuse or by the willingness to
commit crime. Used in the right circumstances though, such approaches may have value.
3.1 Feedback from End User Organisations and Practitioners
The opportunities new technologies present to monitor and prevent domestic abuse were explored during a series of
workshops with academic experts, practitioners, and a selection of engineers and technologists, including representatives
from police forces, government, charities, trusts, voluntary support groups and end users. Few workshops took place
between June 2018 and May 2020. The content of the different workshops is summarised as follows:
1.
Exploratory Workshops (World Cafes): We held two workshops to explore how developing technologies might
help to reduce the risk of domestic abuse. The aim of the first workshop was to work with domain experts
to produce a "requirements brief” that mapped out a handful of priority problems, together with any factors
10
APREPRINT - FE BRUARY 9, 2021
Figure 8: Illustration of proximity detection technology around crowds.
that constrain how they are dealt with now and how they might be dealt with in the future. The aim of the
second workshop was to identify possible solutions to the problems identified. A mixture of domain experts,
designers and engineers were invited to the second workshop.
2.
Co-Design Workshop: Existing examples of crime prevention technologies were discussed to explore the
opportunities they offer. Also, the difficulties that survivors might face in calling for help or recording evidence
discreetly were discussed. Few potential technologies that might help in supporting both victims, police and
aid evidence gathering were introduced. Participants particularly liked the portability of small Edge Computing
platforms including wearable devices and the different methods of interaction as compared to smartphones.
Participants were excited by the concept of proximity detection techniques being able to sense the presence
of individuals and any abnormal activities in the surrounding environment and the nearby vicinity as many
vulnerable individuals might find it difficult to record such incidents. The possibility for IoT devices to provide
evidence and a reliable journal was also intriguing as the participants had not acknowledged these technologies
before.
3.
Co-Creation Workshop: A reduced number of possible technological solutions including tagging and proximity
detection were discussed along with hand-on demonstrations. First the potential use of short range tagging
technology as part of a panic alarm was discussed. The scenario proposed was one where small tags “dots”
are placed in home environments, or in public places, which then can be used to send an alarm or as an
information access point. Proximity detection technologies [
84
,
85
] were also discussed due to their potential
in ensuring regular offenders keep their legal distance from victims. Fig. 8 shows how the alerts can be sent
out depending on the proximity of the devices of interest. An alternative suggestion considered the use of
internet connected devices, such as smart speakers, to automatically detect incidents of abuse whilst these
were in progress. To reduce false alarms, it was suggested that such a system could scan for multiple signals to
include slammed doors, raised voices, particular keywords in verbal exchanges, or other indicators of threats of
risk. We discussed the practicality of such a system and how to protect user’s privacy, as well as how triggering
actions could be tailored to the individuals in question to increase the sensitivity of the system. Once triggered,
data can be stored locally or on the cloud, and alerts can be sent to trusted third parties.
4.
One to one discussions with End Users Organisations: Based on the discussions in the previous workshops,
the aim of these series of online discussions was to narrow down the possible development routes and to come
up with a list of paper prototypes which were discussed with each organisation to understand their implications
and limitations.
5.
Design Evaluation Workshop: In the last of these workshops, the previously discussed technologies narrowed
down the proposed prototypes, including the use of short-range communication mediums for alert and
recording of abnormal activities in close proximity and also the use of AI to detect violent behaviour. During
the workshop the discussion converged on the benefits of a single edge device aimed at crime prevention and
the collection of evidence that would protect users privacy. The use of edge computing would mean that data
would not need to be stored remotely and the available processing power would mean that computational
intense processing (e.g. voice and sentiment detection) could be carried out on such a device. Personalised and
more adaptable interfaces or functionalities of the device then can be developed targeting diverse groups of
people.
The series of co-design and co-creation workshops carried out concluded that there is a need for a range of technological
solutions to address crime prevention, as a one-size-fits-all solution could not feasibly address the different issues and
11
APREPRINT - FE BRUARY 9, 2021
potential users. In addition, a single organisation cannot solve all the problems alone, a collaborative effort is thus
required to reach optimal solutions. One of the workshops participants David Lucy from the Met Police explains:
"The Police Service is facing increased and changing demand, putting pressure on front line services. Powerful new
technologies provide a way for the Police Service to prevent crime and serve their communities more efficiently. Crime
is changing in the digital age. The police service must continue to adapt, to deal with new and emerging cyber-crimes,
whilst still addressing the demand from things such as acquisitive crime and domestic abuse. Demand from both
emerging and existing crimes can be reduced through the use of new technology. We have been engaged in a series of
co-design sessions and sandpits with collaborators from Nottingham Trent University (NTU) and the University College
London (UCL) discussing the potential of smart devices and technologies to prevent crimes and support vulnerable
individuals. By engaging closely in these activities, we have learnt about the potential of proximity detection and
tagging technologies in alerting relevant support organisation or family members to an imminent risk. We firmly
believe these should be further explored and trialed since they show the potential to support policing with preventing
crime, supporting investigations but ultimately achieving the best outcomes in the pursuit of justice and in the support
of victims". The workshops have also helped in identifying and understanding the key research challenges faced by
researchers who work in this research field. Such challenges are summarised in the section below.
4 Research Challenges
This section presents the main research challenges faced at the development of pervasive computing technologies for
preventing criminal activity (see Fig. 9).
4.1 Information Privacy
There has been much research conducted in the recent years in relation to the privacy implications of smart devices.
However, as stated in [
86
], system developers are normally compelled to meet strict deadlines to avoid losing competitive
advantage, leading to immature products which fail to satisfy the expected privacy and security requirements of their
specific target applications [87].
A large number of users may get understandably uncomfortable with sending their personal data to remote data stores
or clouds which they cannot see or control. Transferring user data over networks (including secure networks) make
systems vulnerable to theft and distortion. To be perceived as legitimate by the public, and to meet ethical and legal
standards, privacy requirements and concerns need to be satisfactorily addressed.
The above privacy concerns can often be addressed by guaranteeing that a user’s personal data will be processed at
the point of collection and will not be transferred remotely. By employing edge computing and on-device processing
techniques, chipsets can potentially be exploited at or close to the data source for performing computational tasks
instead of transmitting all the data and processing it in a central server. A competent trade-off between personal privacy
and computational power should certainly be optimised for each crime prevention tool if this is to be widely employed
by the public.
4.2 Diversity and Scalability
Developing AI technologies to detect explicit personal behaviours or actions requires the collection of data from a large
number of users, to enable predictive models incorporate sufficient variability in order to generalise well on unseen
data. Therefore, it is crucial to carry out data collection processes on a large number of experimental participants, as
well as to consider the differences that may exist between different groups of individuals. For instance, let’s imagine
one is planning to develop and launch a novel audio-based system to recognise violent behaviours for the prevention
of domestic abuse in households within the UK. The first problem a system developer may encounter is that not
every household speaks English. A speech recognition model which is only trained with data from native speakers
only, would certainly not perform well with data processed from non-native speakers. Additionally, there are other
sound characteristics, such as tone, phonetics, intonation and melody which may vary considerably between different
languages or even between different accents of the same language. Thus, although it may involve tedious processes
of data collection, it is crucial to identify the target group and incorporate an adequate level of inter and intra-group
variability when developing intelligent systems. This issue becomes even more crucial when dealing with sensitive
matters like crime where false negatives can translate into serious personal health-related consequences.
12
APREPRINT - FE BRUARY 9, 2021
Figure 9: Challenges of adopting technological innovations for Crime prevention.
4.3 Data Integrity
Data integrity refers to the consistency, accuracy and completeness over the entire life cycle of the different data
retrieval, data transmission, data storage and data processing units of a system. Maintaining data integrity is crucial for
various reasons. First, data integrity ensures recover-ability and trace-ability. Second, the decision-making made by
intelligent systems is driven by the data it is developed upon, thus the accuracy of such systems strongly depends on the
integrity of the data they are developed, maintained and functioned with.
4.4 Battery Life
Energy reduction and sustainability have become major issues in the technical and social agendas in the last years. The
ubiquitous and pervasive computing research communities are systematically facing a strain between usability and
sustainability. On the one hand, users express an increasing interest in purchasing more sustainable products. On the
other hand, they do not wish to do so at the sacrifice of their comfort.
The widespread use of mobile and portable devices such as smart-phones or tablets comes alongside an ever increasing
demand for mobile apps, including resource demanding applications such as face, speech or activity recognition.
Although resources like memory, bandwidth and processing power are constantly being improved to keep up with the
increasing demand for additional storage, communication and processing power, battery capacity grows only at an
approximate rate of 10% per year [
88
]. However, increasing battery capacity is not the only way to improve battery
lifetime. Alternatively, research efforts have been and should still be made to explore ways to reduce power consumption
from mobile devices.
4.5 Accuracy and Experimental Constraints
Intuitive optimisation of various parameters often result in better and more accurate models, with error rates being
continuously lowered. However, there exist various limitations and challenges which do not normally allow Artificial
Intelligence (AI) applications to exhibit error-free performances. First, machine learning and deep learning algorithms
typically require large amounts of data to successfully undergo the training phase. In addition, as mentioned in Section
4.3, such data should incorporate the adequate inter and intra-subject variability for the classification or regression
models to be able to generalise well on unseen data samples. This means that in addition to the need to collect large
amounts of data, that data has to be variable enough to adequately represent the characteristics of the target population
and avoid large generalisation or out-of-sample errors. Further to ensuring the collection of "enough" data samples,
adequate feature engineering and machine learning techniques can be a crucial factor to optimise the accuracy achieved
by AI-based crime prevention systems. A research challenge therefore arises from the need to improve the performance
achieved by the state-of-the-art in the corresponding fields. For instance, as a current survey on sentiment analysis
indicates [
89
], the classification accuracy achieved by the state-of-the-art on sentiment analysis using audio recordings,
is in the range of 72.9% to 85.1%. This means that if an audio-based verbal abuse system was to be developed, it would
be wrong 14.9% of the cases. The behavioural data required for these models (e.g. [
90
] for video Processing [
91
] for
13
APREPRINT - FE BRUARY 9, 2021
emotion recognition [
92
,
77
] for activity recognition), need to be collected under highly variant free-living scenarios
rather than a controlled settings.
4.6 Affordability
Cost was a key factor discussed during the focus group as end users and other supporting organisations would need
the device to be as inexpensive as possible if it was to become adopted into practice. In this regard, while tagging
technologies relatively cheap, applications requiring high processing power may depend upon expensive hardware.
Edge computing enables processing all or part of the data at the location it is collected. Data that is only of ephemeral
importance can be crunched on the edge device itself. This is in contrast to cloud-based systems, where data is sent to
large, remote data centers for processing. In accordance with Moore’s Law [
93
], small devices at the edge have become
more computationally powerful. If the trend is to continue, it is only inevitable that switching to edge platforms would
offer much affordable solutions in the long run. This territorial proximity to the endpoint is good for both latency and
efficiency as it saves networks from unnecessary congestion as well as from carrying sensitive personal data.
The ongoing minimisation and mass production of electronics has enabled a reduction in the cost of edge computing
devices, whereby various complex computations (such as AI and data processing) can take place on-board. However, the
more data and computing intensive an application is, the more data storage and processing power are required, with this
having an impact on the price to be paid by end users. In this regard, the adoption of high processing power technologies
by the public has two main research challenges associated to it. First, to keep up with the ongoing miniaturisation and
cost reduction of electronic components. And secondly, the optimisation of signal processing and machine learning
algorithms so these can be adopted utilising a lower computing power.
4.7 Technology Misuse and Technology Literacy
Advancement in technologies enables law enforcement and voluntary services to support victims and reduce or prevent
crimes. However, the increasing complexity and communication capabilities present in these technologies have also
opened new pathways for data interference[94].
There is little empirical research published concerning the use of technology in intimate partner stalking, as most of the
current efforts are focused on online abuse on social media or texting [
95
]. Within these, the work in [
95
] conducted a
survey with 152 domestic violence advocates and 46 victims. The study found that modern technologies can potentially
give perpetrators multiple tools to control and manipulate people and that technology-facilitated stalking needs to
be treated as a serious offense. There is thus a need for non-judgemental responses from service providers and law
enforcement to victims experiencing such abuse. As practitioners observed, advising victims to switch off devices,
to withdraw from social media, or to change their profile or telephone numbers, is putting an enormous burden of
responsibility on the victim to adjust their behaviour [
96
]. Additionally, disengagement from technology can mean
that victims are increasingly uncontactable, which can impact the type and the timing of the support they receive from
services.
There is a great need to increase public awareness of the use of spyware to commit abuse and stalking. Likewise, law
enforcement and victim support and rehabilitation organisations will benefit from learning about the latest development
of technologies that might help vulnerable individuals or prevent technology misuse.
5 Conclusions and Directions for Future Work
Different technology-driven solutions and the potential adoption of contemporary smart pervasive, machine intelligence
systems and miniature technologies for crime prevention have been considered and evaluated. To our knowledge, this is
the first paper that reviews recent technological innovation in crime prevention.
As discussed above, in our view, there is enormous potential associated with the adoption of short-range communication
technologies as a tool for crime prevention. These technologies can be employed alongside current approaches, such
as GPS monitoring, to provide a more robust proximity detection system, able to detect proximity in both indoor and
outdoor environments.
Smart and short-range tags could be employed to provide an instant access to information and also for emergency
reporting. As discussed, several apps to report emergency situations and collect evidence of those situations have been
developed in recent years. However, these apps commonly require the user to open the app and perform an specific
action.
14
APREPRINT - FE BRUARY 9, 2021
Furthermore, violence detection scenarios in home or work environments, the use of a audio-based technologies appear
to be preferred against that of CCTV cameras given the ubiquity, spherical field and lower privacy concerns exhibited by
the former systems. As exposed, the use of audio signal processing along with machine learning techniques, allowing
for applications such as speaker diarization, speech recognition, person identification and sentiment analysis, could be
key to identify violent language as well as violent actions.
From a technical viewpoint, standalone systems employing a single sensing technology exhibit distinct limitations.
However, the combination of technologies, can be of great use to identify violent scenarios, which can lead to the
prevention of further occurrences and therefore to the ultimate prevention of criminal activity. In conclusion, not all
circumstances or situations are the same and there will be no “one size fits all” solution(s). As such, it is important
to explore the heterogeneity in offenders, victims, contexts of offending, and offending patterns to understand which
solutions might work for whom, and under what circumstances. Beside privacy, scalability, affordability, miniaturisation
and personalisation are some of the important factors that need to be considered when designing technologies for crime
prevention.
Nonetheless, future work will embody the conduct of focus groups with co-design and co-creation elements where the
findings and conclusions drawn in this paper will be further discussed and analysed with end user organisations and
various groups of interest. With this, we aim to gain more insights into the potential, limitations and drawbacks of the
discussed technologies and obtain critical advise from experts in the field. The development and implementation of a
number of prototypes of the selected crime prevention systems will follow.
6 Acknowledgement
This research was funded by the Dawes Center for Future Crime at UCL. We would like to thank all the participants in
our workshops for contributing to the discussion and co-design sessions. We would like to thank all the participants in
our workshops who have openly shared their views and ideas. Each of their individual contributions has helped us to
better understand the facets of co-design for crime prevention. The discussions in the workshops have played a pivotal
role in the development of our views on this topic.
References
[1]
Mark Shaw, Jan Van Dijk, and Wolfgang Rhomberg. Determining trends in global crime and justice: An overview
of results from the united nations surveys of crime trends and operations of criminal justice systems. In Forum on
crime and society, volume 3, pages 35–63. Citeseer, 2003.
[2] Meghan Elkin. Crime in england and wales: year ending march 2019, Jul 2019.
[3]
Richard B Felson, Steven F Messner, Anthony W Hoskin, and Glenn Deane. Reasons for reporting and not
reporting domestic violence to the police. Criminology, 40(3):617–648, 2002.
[4] Gill Hague and Ellen Malos. Domestic violence: Action for change. New Clarion Pr, 2005.
[5] Sharon D Herzberger. Violence within the family: Social psychological perspectives. Routledge, 2019.
[6]
Graham Farrell, Nick Tilley, and Andromachi Tseloni. Why the crime drop? Crime and justice, 43(1):421–490,
2014.
[7] Nick Stripe. Crime in england and wales: year ending june 2020, Oct 2020.
[8]
Janet BL Chan. The technological game: How information technology is transforming police practice. Criminal
justice, 1(2):139–159, 2001.
[9]
C Harris. Police and soft technology: How information technology contributes to police decision making. The
new technology of crime, law and social control, pages 153–183, 2007.
[10]
M Stroshine. Information technology innovations in policing. Critical Issues in Policing: Contemporary Readings,
pages 172–183, 2005.
[11]
Zhenyu Chen, Yiqiang Chen, Xingyu Gao, Shuangquan Wang, Lisha Hu, Chenggang Clarence Yan, Nicholas D
Lane, and Chunyan Miao. Unobtrusive sensing incremental social contexts using fuzzy class incremental learning.
In 2015 IEEE International Conference on Data Mining, pages 71–80. IEEE, 2015.
[12]
James Byrne and Gary Marx. Technological innovations in crime prevention and policing. a review of the research
on implementation and impact. Journal of Police Studies, 20(3):17–40, 2011.
[13]
Eric L Piza, Brandon C Welsh, David P Farrington, and Amanda L Thomas. Cctv surveillance for crime prevention:
A 40-year systematic review with meta-analysis. Criminology & Public Policy, 18(1):135–159, 2019.
15
APREPRINT - FE BRUARY 9, 2021
[14]
Clive Norris and Michael McCahill. Cctv: Beyond penal modernism? British Journal of Criminology, 46(1):97–
118, 2005.
[15]
Joshua C Klontz and Anil K Jain. A case study of automated face recognition: The boston marathon bombings
suspects. Computer, 46(11):91–94, 2013.
[16]
Nurul Azma Abdullah, Md Jamri Saidi, Nurul Hidayah Ab Rahman, Chuah Chai Wen, and Isredza Rahmi A
Hamid. Face recognition for criminal identification: An implementation of principal component analysis for face
recognition. In AIP Conference Proceedings, volume 1891, page 020002. AIP Publishing, 2017.
[17]
Imed Bouchrika, John N Carter, and Mark S Nixon. Towards automated visual surveillance using gait for
identity recognition and tracking across multiple non-intersecting cameras. Multimedia Tools and Applications,
75(2):1201–1221, 2016.
[18]
Wei Zeng and Cong Wang. View-invariant gait recognition via deterministic learning. Neurocomputing, 175:324–
335, 2016.
[19]
Medha Bhargava, Chia-Chih Chen, Michael S Ryoo, and Jake K Aggarwal. Detection of abandoned objects in
crowded environments. In 2007 IEEE Conference on Advanced Video and Signal Based Surveillance, pages
271–276. IEEE, 2007.
[20]
Diego Gragnaniello, Carlo Sansone, and Luisa Verdoliva. Iris liveness detection for mobile devices based on local
descriptors. Pattern Recognition Letters, 57:81–87, 2015.
[21]
Mahasak Ketcham, Thittaporn Ganokratanaa, and Sriphagaarucht Srinhichaarnun. The intruder detection system
for rapid transit using cctv surveillance based on histogram shapes. In 2014 11th International Joint Conference
on Computer Science and Software Engineering (JCSSE), pages 1–6. IEEE, 2014.
[22]
Seung Hyun Kim, Su Chang Lim, et al. Intelligent intrusion detection system featuring a virtual fence, active
intruder detection, classification, tracking, and action recognition. Annals of Nuclear Energy, 112:845–855, 2018.
[23]
Rutvik Kakadiya, Reuel Lemos, Sebin Mangalan, Meghna Pillai, and Sneha Nikam. Ai based automatic
robbery/theft detection using smart surveillance in banks. In 2019 3rd International conference on Electronics,
Communication and Aerospace Technology (ICECA), pages 201–204. IEEE, 2019.
[24]
Teng Li, Huan Chang, Meng Wang, Bingbing Ni, Richang Hong, and Shuicheng Yan. Crowded scene analysis: A
survey. IEEE transactions on circuits and systems for video technology, 25(3):367–386, 2014.
[25]
William D Burrell and Robert S Gable. From bf skinner to spiderman to martha stewart: The past, present and
future of electronic monitoring of offenders. Journal of Offender Rehabilitation, 46(3-4):101–118, 2008.
[26]
Jo Brayford, Francis Cowe, and John Deering. Sex Offenders: Punish, Help, Change Or Control?: Theory, Policy
and Practice Explored. Routledge, 2013.
[27]
Mary A Finn and Suzanne Muirhead-Steves. The effectiveness of electronic monitoring with violent male parolees.
Justice Quarterly, 19(2):293–312, 2002.
[28]
Edna Erez and Peter R Ibarra. Making your home a shelter: Electronic monitoring and victim re-entry in domestic
violence cases. British journal of criminology, 47(1):100–120, 2006.
[29]
Eric Grommon, Jason Rydberg, and Jeremy G Carter. Does gps supervision of intimate partner violence defendants
reduce pretrial misconduct? evidence from a quasi-experimental study. Journal of Experimental Criminology,
13(4):483–504, 2017.
[30]
Hugh Downing. The emergence of global positioning satellite (gps) systems in correctional applications. Correc-
tions Today, 68(6):42, 2006.
[31]
Guy Griffiths, Shane D Johnson, and Kevin Chetty. Uk-based terrorists’ antecedent behavior: A spatial and
temporal analysis. Applied geography, 86:274–282, 2017.
[32]
Jyoti Belur, Amy Thornton, Lisa Tompson, Matthew Manning, Aiden Sidebottom, and Kate Bowers. A systematic
review of the effectiveness of the electronic monitoring of offenders. Journal of Criminal Justice, 68:101686,
2020.
[33]
Karl Benkic, Marko Malajner, P Planinsic, and Z Cucej. Using rssi value for distance estimation in wireless sensor
networks based on zigbee. In 2008 15th International Conference on Systems, Signals and Image Processing,
pages 303–306. IEEE, 2008.
[34]
Omotayo G Adewumi, Karim Djouani, and Anish M Kurien. Rssi based indoor and outdoor distance estimation
for localization in wsn. In 2013 IEEE international conference on Industrial technology (ICIT), pages 1534–1539.
IEEE, 2013.
16
APREPRINT - FE BRUARY 9, 2021
[35]
Abdalkarim Awad, Thorsten Frunzke, and Falko Dressler. Adaptive distance estimation and localization in wsn
using rssi measures. In 10th Euromicro Conference on Digital System Design Architectures, Methods and Tools
(DSD 2007), pages 471–478. IEEE, 2007.
[36]
Eamonn O’Neill, Vassilis Kostakos, Tim Kindberg, Alan Penn, Danaë Stanton Fraser, Tim Jones, et al. Instru-
menting the city: Developing methods for observing and understanding the digital cityscape. In International
Conference on Ubiquitous Computing, pages 315–332. Springer, 2006.
[37]
Tom Nicolai and Holger Kenn. About the relationship between people and discoverable bluetooth devices in
urban environments. In Proceedings of the 4th international conference on mobile technology, applications, and
systems and the 1st international symposium on Computer human interaction in mobile technology, pages 72–78.
ACM, 2007.
[38]
Ryo Nishide and Hideyuki Takada. Detecting pedestrian flows on a mobile ad hoc network and issues with trends
and feasible applications. International Journal on Advances in Networks and Services, 6(1&2), 2013.
[39]
Clifford S Fishman. The interception of communications without a court order: Title iii, consent, and the
expectation of privacy. . John’s L. Rev., 51:41, 1976.
[40]
Daniel S Lawrence, Nancy G La Vigne, Margaret Goff, and Paige S Thompson. Lessons learned implementing
gunshot detection technology: Results of a process evaluation in three major cities. Justice Evaluation Journal,
1(2):109–129, 2018.
[41]
Hyunji Chung, Michaela Iorga, Jeffrey Voas, and Sangjin Lee. Alexa, can i trust you? Computer, 50(9):100–104,
2017.
[42]
Xavier Anguera, Simon Bozonnet, Nicholas Evans, Corinne Fredouille, Gerald Friedland, and Oriol Vinyals.
Speaker diarization: A review of recent research. IEEE Transactions on Audio, Speech, and Language Processing,
20(2):356–370, 2012.
[43]
Jordan J Bird, Anikó Ekárt, and Diego R Faria. Learning from interaction: An intelligent networked-based
human-bot and bot-bot chatbot system. In UK Workshop on Computational Intelligence, pages 179–190. Springer,
2018.
[44]
Fred Richardson, Douglas Reynolds, and Najim Dehak. Deep neural network approaches to speaker and language
recognition. IEEE signal processing letters, 22(10):1671–1675, 2015.
[45]
David Snyder, Daniel Garcia-Romero, Gregory Sell, Daniel Povey, and Sanjeev Khudanpur. X-vectors: Robust
dnn embeddings for speaker recognition. In 2018 IEEE International Conference on Acoustics, Speech and Signal
Processing (ICASSP), pages 5329–5333. IEEE, 2018.
[46]
Rosalind W Picard. Automating the recognition of stress and emotion: From lab to real-world impact. IEEE
MultiMedia, 23(3):3–7, 2016.
[47]
Akane Sano and Rosalind W Picard. Stress recognition using wearable sensors and mobile phones. In 2013
Humaine Association Conference on Affective Computing and Intelligent Interaction, pages 671–676. IEEE, 2013.
[48]
Eiman Kanjo, Luluah Al-Husain, and Alan Chamberlain. Emotions in context: examining pervasive affective
sensing systems, applications, and analyses. Personal and Ubiquitous Computing, 19(7):1197–1212, 2015.
[49]
Yekta Said Can, Bert Arnrich, and Cem Ersoy. Stress detection in daily life scenarios using smart phones and
wearable sensors: A survey. Journal of biomedical informatics, page 103139, 2019.
[50]
Simon Ollander, Christelle Godin, Sylvie Charbonnier, and Aurélie Campagne. Feature and sensor selection for
detection of driver stress. In PhyCS, pages 115–122, 2016.
[51]
Yun Liu and Siqing Du. Psychological stress level detection based on electrodermal activity. Behavioural brain
research, 341:50–53, 2018.
[52]
Jennifer A Healey and Rosalind W Picard. Detecting stress during real-world driving tasks using physiological
sensors. IEEE Transactions on intelligent transportation systems, 6(2):156–166, 2005.
[53]
Mark Colvin, Francis T Cullen, and Thomas Vander Ven. Coercion, social support, and crime: An emerging
theoretical consensus. Criminology, 40(1):19–42, 2002.
[54] Sheilagh Hodgins. Mental disorder, intellectual deficiency, and crime: evidence from a birth cohort. Archives of
general psychiatry, 49(6):476–483, 1992.
[55]
Ann L Coker, Paige H Smith, Lesa Bethea, Melissa R King, and Robert E McKeown. Physical health consequences
of physical and psychological intimate partner violence. Archives of family medicine, 9(5):451–457, 2000.
[56]
Loring Jones, Margaret Hughes, and Ulrike Unterstaller. Post-traumatic stress disorder (ptsd) in victims of
domestic violence: A review of the research. Trauma, Violence, & Abuse, 2(2):99–119, 2001.
17
APREPRINT - FE BRUARY 9, 2021
[57]
Donna Eileen Stewart and Simone Natalie Vigod. Mental health aspects of intimate partner violence. Psychiatric
Clinics, 40(2):321–334, 2017.
[58]
Oliver Hutt, Kate Bowers, Shane Johnson, and Toby Davies. Data and evidence challenges facing place-based
policing. Policing: An International Journal, 41(3):339–351, 2018.
[59] Rachel Boba Santos. Crime analysis with crime mapping. Sage publications, 2016.
[60]
Adam A Abbas, Abubakar U Alhaji, and Kadini J Bitrus. Locational analysis of crime in gombe metropolis,
nigeria. 2017.
[61]
Shane D Johnson. A brief history of the analysis of crime concentration. European Journal of Applied Mathematics,
21(4-5):349–370, 2010.
[62]
Anthony A Braga, Martin A Andresen, and Brian Lawton. The law of crime concentration at places: Editors’
introduction, 2017.
[63]
Simon Williams and Timothy Coupe. Frequency vs. length of hot spots patrols: a randomised controlled trial.
Cambridge Journal of Evidence-Based Policing, 1(1):5–21, 2017.
[64]
J Byrne. The new generation of concentrated community supervision strategies: Focusing resources on high risk
offenders, times, and places. A Report for the Public Safety Performance Project, the Pew Charitable Trusts:
Washington, DC, 2009.
[65] Christopher Soghoian. The law enforcement surveillance reporting gap. Available at SSRN 1806628, 2011.
[66] Bill Fisher. Uk time spent with media 2019, May 2019.
[67] Circle of 6. Circle of 6, 2015 (accessed October 12, 2020).
[68] Noonlight. Noonlight, 2020 (accessed October 13, 2020).
[69] Silent Beacon. Silent Beacon, 2019 (accessed October 12, 2020).
[70] Red Panic Button. Red Panic Button, 2020 (accessed October 14, 2020).
[71] bSafe. bSafe, 2020 (accessed October 14, 2020).
[72] Bright Sky. Bright Sky, 2020 (accessed October 14, 2020).
[73] SpotCrime+ Crime Map. SpotCrime+ Crime Map, 2020 (accessed October 15, 2020).
[74] Henry Williams. Police Crime Statistics Map, 2020 (accessed October 15, 2020).
[75] Miranda Warnings / Rights. Miranda Warnings / Rights, 2020 (accessed October 15, 2020).
[76]
Julio Suarez-Paez, Mayra Salcedo-Gonzalez, Alfonso Climente, Manuel Esteve, Jon Ander Gómez, Carlos Enrique
Palau, and Israel Pérez-Llopis. A novel low processing time system for criminal activities detection applied to
command and control citizen security centers. Information, 10(12):365, Nov 2019.
[77]
Dario Ortega Anderez, Ahmad Lotfi, and Amir Pourabdollah. Eating and drinking gesture spotting and recog-
nition using a novel adaptive segmentation technique and a gesture discrepancy measure. Expert Systems with
Applications, 140:112888, 2020.
[78]
Jerry H Ratcliffe, Matthew Lattanzio, George Kikuchi, and Kevin Thomas. A partially randomized field experiment
on the effect of an acoustic gunshot detection system on police incident reports. Journal of Experimental
Criminology, 15(1):67–76, 2019.
[79]
Pierre Laffitte, Yun Wang, David Sodoyer, and Laurent Girin. Assessing the performances of different neural
network architectures for the detection of screams and shouts in public transportation. Expert Systems With
Applications, 117:29–41, 2019.
[80]
Jianfeng Chen, Alvin Harvey Kam, Jianmin Zhang, Ning Liu, and Louis Shue. Bathroom activity monitoring
based on sound. In International Conference on Pervasive Computing, pages 47–61. Springer, 2005.
[81]
Tara Matthews, Kathleen O’Leary, Anna Turner, Manya Sleeper, Jill Palzkill Woelfer, Martin Shelton, Cori
Manthorne, Elizabeth F Churchill, and Sunny Consolvo. Stories from survivors: Privacy & security practices
when coping with intimate partner abuse. In Proceedings of the 2017 CHI Conference on Human Factors in
Computing Systems, pages 2189–2201. ACM, 2017.
[82]
John Eck, Spencer Chainey, James Cameron, and Ronald Wilson. Mapping crime: Understanding hotspots. 2005.
[83]
Emily Van der Nagel and Jordan Frith. Anonymity, pseudonymity, and the agency of online identity: Examining
the social practices of r/gonewild. First Monday, 20(3), 2015.
18
APREPRINT - FE BRUARY 9, 2021
[84]
Kieran Woodward, Eiman Kanjo, Dario Ortega Anderez, Amna Anwar, Thomas Johnson, and John Hunt.
Digitalppe: low cost wearable that acts as a social distancingreminder and contact tracer. In Proceedings of the
18th Conference on Embedded Networked Sensor Systems, pages 758–759, 2020.
[85]
Dario Ortega Anderez, Eiman Kanjo, Ganna Pogrebna, Omprakash Kaiwartya, Shane D Johnson, and John Alan
Hunt. A covid-19-based modified epidemiological model and technological approaches to help vulnerable
individuals emerge from the lockdown in the uk. Sensors, 20(17):4967, 2020.
[86]
Hadi Habibzadeh, Brian H Nussbaum, Fazel Anjomshoa, Burak Kantarci, and Tolga Soyata. A survey on
cybersecurity, data privacy, and policy issues in cyber-physical system deployments in smart cities. Sustainable
Cities and Society, 2019.
[87]
Rida Khatoun and Sherali Zeadally. Cybersecurity and privacy solutions in smart cities. IEEE Communications
Magazine, 55(3):51–59, 2017.
[88]
Gabriel Orsini, Dirk Bade, and Winfried Lamersdorf. Cloudaware: A context-adaptive middleware for mobile edge
and cloud computing applications. In 2016 IEEE 1st International Workshops on Foundations and Applications of
Self* Systems (FAS* W), pages 216–221. IEEE, 2016.
[89]
Nusrat J Shoumy, Li-Minn Ang, Kah Phooi Seng, DM Motiur Rahaman, and Tanveer Zia. Multimodal big
data affective analytics: A comprehensive survey using text, audio, visual and physiological signals. Journal of
Network and Computer Applications, 149:102447, 2020.
[90]
Jordan J Bird, Elizabeth Wanner, Anikó Ekárt, and Diego R Faria. Accent classification in human speech
biometrics for native and non-native english speakers. In PETRA, pages 554–560, 2019.
[91]
Eiman Kanjo, Eman MG Younis, and Chee Siang Ang. Deep learning analysis of mobile physiological, environ-
mental and location sensor data for emotion detection. Information Fusion, 49:46–56, 2019.
[92]
Dario Ortega-Anderez, Ahmad Lotfi, Caroline Langensiepen, and Kofi Appiah. A multi-level refinement approach
towards the classification of quotidian activities using accelerometer data. Journal of Ambient Intelligence and
Humanized Computing, 10(11):4319–4330, 2019.
[93] John L. Gustafson. Moore’s Law, pages 1177–1184. Springer US, Boston, MA, 2011.
[94]
Diana Freed, Jackeline Palmer, Diana Minchala, Karen Levy, Thomas Ristenpart, and Nicola Dell. “a stalker’s
paradise”: How intimate partner abusers exploit technology. In Proceedings of the 2018 CHI Conference on Human
Factors in Computing Systems, CHI ’18, page 1–13, New York, NY, USA, 2018. Association for Computing
Machinery.
[95]
Delanie Woodlock. The abuse of technology in domestic violence and stalking. Violence Against Women,
23(5):584–602, 2017.
[96]
Delanie Woodlock, Mandy McKenzie, Deborah Western, and Bridget Harris. Technology as a weapon in domestic
violence: Responding to digital coercive control. Australian Social Work, 73(3):368–380, 2020.
19