Article 11

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

The current issue and full text archive of this journal is available on Emerald Insight at:

https://www.emerald.com/insight/0368-492X.htm

Social media users and Hybrid


artificial
cybersecurity awareness: intelligence
approach
predicting self-disclosure using
a hybrid artificial 401
intelligence approach Received 8 May 2021
Revised 18 July 2021
Accepted 28 August 2021
Naurin Farooq Khan, Naveed Ikram and Hajra Murtaza
Faculty of Computing, Riphah International University, Islamabad, Pakistan, and
Muhammad Aslam Asadi
Bahauddin Zakariya University, Multan, Pakistan

Abstract
Purpose – This study aims to investigate the cybersecurity awareness manifested as protective behavior to
explain self-disclosure in social networking sites. The disclosure of information about oneself is associated with
benefits as well as privacy risks. The individuals self-disclose to gain social capital and display protective
behaviors to evade privacy risks by careful cost-benefit calculation of disclosing information.
Design/methodology/approach – This study explores the role of cyber protection behavior in predicting
self-disclosure along with demographics (age and gender) and digital divide (frequency of Internet access)
variables by conducting a face-to-face survey. Data were collected from 284 participants. The model is
validated by using multiple hierarchal regression along with the artificial intelligence approach.
Findings – The results revealed that cyber protection behavior significantly explains the variance in self-
disclosure behavior. The complementary use of five machine learning (ML) algorithms further validated the
model. The ML algorithms predicted self-disclosure with an area under the curve of 0.74 and an F1 measure
of 0.70.
Practical implications – The findings suggest that costs associated with self-disclosure can be mitigated by
educating the individuals to heighten their cybersecurity awareness through cybersecurity training programs.
Originality/value – This study uses a hybrid approach to assess the influence of cyber protection behavior on
self-disclosure using expectant valence theory (EVT).
Keywords Self-disclosure, Social networking sites, Cybersecurity awareness, Cyber protection behavior,
Artificial intelligence
Paper type Research paper

1. Introduction
The development of information and communication technologies (ICTs) has opened doors to
social and economic opportunities. Socially connecting with others on social networking sites
(SNSs) is now an important part of one’s life with self-disclosure becoming an international
phenomenon. Self-disclosure has been defined as revealing personal information to others to
which individuals already have no access to Wheeless and Grotz (1976). Although self-
disclosure is an old phenomenon, online disclosure of the information is different due to the
disinhibition effect and the speed at which information can be shared publicly (Mubarak and
Rahamathulla, 2016). The online self-disclosure is the voluntary disclosure of information to
others on online platforms (Pujazon-Zazik and Park, 2010).
A burgeoning body of research has been carried out to ascertain the determinants of self- Kybernetes
disclosure (Walsh et al., 2020). The use of SNSs to disclose information is associated with Vol. 52 No. 1, 2023
pp. 401-421
capital-enhancing activities (Aharony, 2016; Tzortzaki et al., 2016). According to Hallam and © Emerald Publishing Limited
0368-492X
Zanella (2017), self-disclosure is considered an integral part of enhancing social capital by the DOI 10.1108/K-05-2021-0377
K creation and maintenance of connections between users. As evidenced in the literature, these
52,1 sites provide platforms and tools for building relationships (Cheung et al., 2015). Users
experience feeling pleasant and having a sense of enjoyment by forming connections
(Chennamaneni and Taneja, 2015). As per Grieve et al. (2013), individuals get relief from
feelings of loneliness by being connected to others. On the other hand, other factors such as
entertainment, self-presentation, attention-seeking and popularity are also important
determinants of self-disclosure behavior, as reported by Abramova et al. (2017). Therefore,
402 there is a positive outcome to self-disclosure that stems from building relationships and
deriving social capital on these SNSs.
Despite the benefits, there are adverse consequences of sharing information on SNSs
(Rauhala et al., 2019) due to easy access to information and privacy violations. SNSs are
designed for easy information upload. This information contains photos, names and
addresses which can make individuals easy targets for abuse (Steijn, 2014). As per Dhir et al.
(2021), maladaptive information-seeking results in compulsive abusive behavior such as
social media stalking. Moreover, according to Salo et al. (2018), such behavior is the
manifestation of voyeuristic motives. This shows that cybercriminals can use SNSs
platforms to collect more and more personal information and allure young individuals into
sex solicitation. The information on SNS platforms can also be used by peers for online
harassment and cyberbullying. According to Gahagan et al. (2016), there is a positive
relationship between self-disclosure and cyberbullying in young individuals. On the other
hand, privacy violations occur when the information is distributed, used or even accessed
without an individual’s consent (Aivazpour and Rao, 2020). In reference to Gharibi and
Shaabi (2012), information released on these platforms increases users’ vulnerabilities and
poses cyber threats to them. Cybercriminals can glean information on SNSs and steal the
identity of the individuals to commit fraud or damage reputation as reported by Loiacono
(2015) and Robinson (2017). Moreover, this information can be used by social engineers to
craft fake emails that can lead to sensitive information leaks such as bank details and
eventual financial loss (Mouton et al., 2016). Anecdotal evidence suggests that information
posted online has also been used to conduct kidnappings and blackmailing (Facebook et al.,
2020). In some extreme cases, individuals have committed suicides due to consistent
blackmailing fearing the broadcasting of their intimate pictures among social networking
contact lists (“Webcam blackmail cases have doubled, police say,” 2016). Crimes associated
with self-disclosure not only affect individuals but also organizations. According to Kefi and
Perez (2018), hackers can easily carry out their malicious activities using information from
SNSs. Such activities result in compromised personal accounts that eventually lead to data
security leakages at the organizational level (Ajis and Salleh, 2020). For instance, LinkedIn
had its customer’s data leaked by hackers who used information from SNSs and further
aggregated it from different websites in 2017 and 2021 (Segura, 2017; “LinkedIn Data Breach -
500M Records Leaked and Being Sold,” 2021). Similarly, earlier in 2021, hackers
automatically harvested data from Facebook accounts. The data of 533 million users was
published on a local website (Haskell-Dowland, n.d.), thereby resulting in privacy violations
of Facebook users. This suggests negative implications of self-disclosure at an individual and
corporate level.
With both positive impacts and negative consequences, self-disclosure acts as a double-
edged sword. Therefore, individuals are concerned about the privacy and security of their
information (Heravi et al., 2018). The literature reports individuals use cost mitigating
factors to manifest these concerns in order to protect against cybercrimes. According to
Stutzman et al. (2011), users make use of privacy settings in SNSs to limit the visibility of
their profile and interaction with different individuals. Li et al. (2019) have found that SNSs
users calculate the risk-benefits trade-off by adopting privacy settings or technical
controls. These controls allow for the users to mitigate privacy-related costs of self-
disclosure while maintaining its benefits ensuing from the capital-enhancing activities on Hybrid
the SNS. As a result, the cost mitigating factors are found to have a positive influence on artificial
self-disclosure (Abramova et al., 2017). Studies have shown that the use of technical
safeguards is an important facilitating factor for the disclosure of information (Abramova
intelligence
et al., 2017; Hajli and Lin, 2016). However, most of the previous research studies consider approach
such technical safeguards in the light of privacy, privacy concerns and associated behavior
with inconsistent definitions (Parsons et al., 2016).
One aspect of such inconsistencies is that the concept of privacy is considered 403
multidimensional in nature that encompasses protective behaviors in terms of not only
privacy but also security (Parsons et al., 2016). A recent study has corroborated these findings
by reporting a correlation between privacy and cybersecurity (Egelman and Peer, 2015a). The
study finds the two concepts to be different due to the orthogonal correlation between them.
This means that the protective behavior to safeguard one’s privacy is not just confined to the
privacy controls present on SNS rather they are also manifested in cyber protection behaviors
such as clearing one’s browser history or using encryption and other cybersecurity measures
such as securing device (Parsons et al., 2016; Ampong et al., 2018). These cyber protection
behaviors come under the umbrella of cybersecurity awareness of the individuals.
Cybersecurity awareness is defined as “The degree of understanding of users about the
importance of information security and their responsibilities and acts to exercise sufficient
levels of information security control to protect the organization’s data and networks” (Shaw
et al., 2009). Another aspect associated with the inconsistencies is that the privacy controls on
SNSs have inflexible privacy settings. The concept of privacy is whitewashed on SNSs and
users do not have complete control of their privacy that these sites claim to provide them
(Barrett-Maitland et al., 2016). Researchers have emphasized this point by showing the
unimportance of privacy due to the norm of personal data collection via targeted ads (Young
and Quan-Haase, 2013). Then there are inherent privacy and security loopholes within SNSs
as reported by Barrett-Maitland et al. (2016). SNSs are not designed with privacy
considerations because this is not the primary focus of an SNS design (Barrett-Maitland
et al., 2016).
This multi-dimensional nature of privacy along with the inadequacy of SNSs’ privacy
controls calls for further scholarship, considering the cybersecurity awareness separately
from the privacy behavior and its influence on users’ self-disclosure. It is only recently that
cybersecurity has been studied in explaining information disclosure. The findings of a recent
study by Rauhala et al. (2019) have emphasized the connection between concerns of negative
consequences and the use of cybersecurity to mediate information disclosure on the Internet.
In a similar vein, another study reports the influence of users who possess the technical
knowledge to secure their mobile devices tend to download mobile apps that ask for excessive
permissions (Barth et al., 2019). Although these two studies take into consideration the
cybersecurity aspect, there are few limitations. The first study (Rauhala et al., 2019) does not
take into account the cybersecurity awareness of the individuals in terms of cyber protection
behavior but takes spending on the software that secures user’s devices to study information
disclosure without any theoretical underpinnings. Whereas, the second study (Barth et al.,
2019) does not use the technical security knowledge of the individuals exclusively to
understand the self-disclosure. Moreover, the study lacks to provide a theoretical framework
in terms of the rational process of cost-benefit analysis. With little to no empirical evidence of
the effects of cyber protection strategies on self-disclosure in SNSs, this study takes into
account the cybersecurity awareness of the users. This behavior is enacted by the users as
protective safeguards – that is a cost mitigating control – to enjoy the capital-enhancing
consequences of the self-disclosure.
K Therefore, this study aims to understand the role of the cyber protection behavior of users
52,1 in influencing their self-disclosure by making use of inferential statistics and artificial
intelligence (AI).

2. Theory and hypothesis development


2.1 Expectant valence theory
404 We take into consideration the expectant valence theory (EVT) to explain the impact of
cybersecurity awareness on self-disclosure in SNSs. Recent literature has suggested that the
EVT provides a flexible foundation in an examination of self-disclosure in the e-commerce
field (Barth and De Jong, 2017; Wilson and Valacich, 2012). The EVT (Vroom, 1964) posits
that the individual’s decision to choose a behavior over another happens due to the expected
result in which benefits are maximized and costs are minimized. The subjective belief system
entails three things: (1) the temptation of receiving a reward and the emotional attachment
toward the rewarding outcome called valence, (2) expectancy – the self-confidence to do
something and (3) likelihood that the reward will be received by the individual called
instrumentality (Barth and De Jong, 2017). Other studies carried out using the theory have
found a significant relationship between perceived security and online shopping behaviors
(Alshare et al., 2019). The EVT has been used to explain the likelihood of online shopping with
the valence of the perceived security, privacy and trust (Alshare et al., 2019). We make use of
the self-confidence (expectancy) component of the EVT to explain self-disclosure that gives a
flexible ground to study the influence of the users’ cybersecurity awareness proficiency.

2.2 Artificial intelligence-based analysis


The use of Artificial Intelligence (AI) is developing rapidly (Gaggioli, 2017) with its
applications spanning healthcare (Wiederhold, 2019a) including psychological research and
human behaviors (Gaggioli, 2017). The AI-based analysis is based on predictions and can be
used to find the generalizability of results as patterns (Bzdok et al., 2018). This predictive
nature of AI can augment statistical analysis which only offers inferences on the data (Bzdok
et al., 2018). It is reported that as compared to indirect assessment criteria, AI-based analysis
allows for high accuracy in predictions (Biran and McKeown, 2017). Therefore, the use of
statistical methods should be augmented by AI-based analysis. Machine learning (ML) can be
used to predict behavior by classifying the data into several predefined classes with respect to
predictive features (Arpaci, 2020). Many techniques are available for carrying out prediction
such as Naı€ve Bayes (NB), Decision Trees (DT), Support Vector Machine (SVM), Logistic
Regression (LR) and K-Nearest Neighbor (KNN). We make use of these algorithms to further
enhance our statistical analysis. Table 1 shows the detail about how these algorithms work
along with the reason for their selection in this study for AI-based analysis.

2.3 Self-disclosure and cyber protection behavior


It is only recently that cybersecurity has been studied in explaining information disclosure.
The findings of a recent study (Rauhala et al., 2019) have recently emphasized the use of
cybersecurity software to mediate information disclosure on the Internet. Similarly, another
study has found that securing mobile devices leads to the disclosure of information while
downloading apps (Barth et al., 2019). Other studies report a similar positive relationship of
perceived security (Gupta and Dhami, 2015) and security awareness (Singh and Gill, 2015)
with disclosing information.
In the light of EVT, the instrumentality, i.e. the likelihood of receiving a reward is already
covered in the action of the self-disclosure since the use of SNSs is simple, and individuals find
it very easy to disclose information. The self-confidence of the users is gained through
Classification
Hybrid
algorithm Working Suitability for the present scenario artificial
intelligence
Logistic Calculates the probabilities of the possible Well suited for binary classification with
Regression (LR) classes using the logistic function discrete features approach
Decision Tree (DT) Decision tree infers a sequence of decision DT produces an interpretable model
rules based upon feature values to classify structure that gives better insight into the
an instance data 405
K-Nearest Computes the label from a simple majority Suited equally to linear and non-linear
Neighbor (K-NN) vote of the k nearest neighbors of the decision boundaries and hence
given data instance compensates for the limitation of LR and
NB
Naı€ve Bayes (NB) NB is based on Bayes’ theorem and Known to work well for simple
calculates the conditional probabilities of classification tasks with small datasets
the possible outcomes with the and is generally used as a baseline to
assumption of independence between compare the performances of other
every pair of features classifiers Table 1.
Support Vector SVM attempts to find the widest possible SVM is scalable in terms of features and Classification
Machines (SVM) class boundaries that separate instances instances and is expected to perform algorithms, their
belonging to different classes in the data consistently in future when the dataset working and reason for
space expands choosing

practicing cyber protection behavior. The valence (EVT component) of the user is associated
with the usage of SNSs for further self-disclosure. The users’ value or valence will perceive the
future event as favorable or unfavorable (Alshare et al., 2019). This study takes into
consideration the value/valence (EVT component) of cyber protection behavior to determine
the users’ choice to self-disclose on SNSs. The users minimize the costs associated with self-
disclosure by their self-confidence in practicing good security behavior to gain the benefits of
disclosing information on social networks. The valence of gaining the associated benefits
of self-disclosures favors information disclosure of the users with the enhanced likelihood of
receiving those benefits and is influenced by the self-confidence emanating from cyber
protection behavior. Therefore, the following hypothesis has been proposed:
H1. Cyber protection behavior positively influences the self-disclosure on SNSs (Figure 1)

Figure 1.
Hypothesis model
K 2.4 Demographics and self-disclosure
52,1 We study the influence of demographics on self-disclosure based on previous literature.
Studies of gender differences in self-disclosure have revealed that the tendency of both
genders to disclose information on SNSs parallels with the face-to-face disclosure
(Almakrami, 2015; Al-Saggaf, 2016; Bond, 2009; Hoy and Milne, 2010; Sheldon, 2013;
Wang et al., 2016). On the other hand, age has been associated with lower self-disclosures,
younger users (Malhotra et al., 2004) are found to self-disclose more compared to older users.
406 Hence, the following hypotheses have been proposed:
H2. Gender influences the self-disclosure on SNSs.
H3. Age negatively influences self-disclosure on SNSs.

2.5 Digital divide and self-disclosure


The digital divide exists among individuals as well as among nations. Literature reports that
once individuals cross the access divide threshold, differences still exist that are associated
with Internet usage (Van Dijk, 2005). The usage of the Internet affects the welfare of
individuals and their online behaviors (Van Deursen et al., 2017). Therefore, the frequency of
Internet access has an effect on self-disclosure habits of individuals. The amount of time the
users spend on the Internet and the related activities are associated with increased self-
disclosure (Shin and Kang, 2016; Xie and Kang, 2015). According to Van Deursen et al. (2017)
and van Ingen and Matzat (2018), this is due to its capital-preserving consequences. The
frequent connectivity leads to more disclosure as per the reciprocal effect. That means the
frequency of Internet access leads to self-disclosure as a result of which the user experiences
enhanced social capital which then leads to further self-disclosure (Shin et al., 2012; Trepte
and Reinecke, 2013). Other studies have found relationships between the frequency of usage
and the increased tendency toward sharing of information on SNSs (Błachnio et al., 2016).
Therefore, the following hypothesis has been proposed:
H4. Frequency of Internet access positively influences self-disclosure.

3. Method
This study uses quantitative research approach. Survey methodology was used to test the
hypothesis. The survey method is useful for understanding the individual self-reported
behaviors by collecting information on a larger scale. The questionnaires were adapted from
the already published literature.

3.1 Measures
The dependent variable in this study is the self-disclosure of users which is measured by the
Personal Information Sharing Practices scale (PISP). The independent variables are cyber
protection behavior, digital divide measured as the frequency of Internet access and
demographics such as gender and age.
3.1.1 Personal Information Sharing Practices (PISP). The PISP measures self-disclosure of
personally identifiable information on SNS. It is a 12 dichotomous item scale which measures
the user’s behavior related to the sharing of individual-specific personally identifiable
information. For example, “Do you include your phone number on your profile?” is coded as
0 if the participant responded as “No” and coded as 1 if the participant responded as “Yes.”
The composite score of PISP scale is derived by adding all the items. The maximum PISP
score is 12 while the minimum is 0. The scale was used by Ball et al. (2015) which adapted it
from (Anderson et al., 2008) and (Furnell, 2008) following the guidelines of Fogel and Nehmad Hybrid
(2009). Table 3 shows the items of the PSIP scale. artificial
3.1.2 Security Behavior Intention Scale (SeBIS). The cyber protection behavior of the
individuals was measured by using Security Behavior Intention Scale (SeBIS). It is a validated
intelligence
scale for measuring the computer security attitudes and related behavior of the end-users approach
(Egelman and Peer, 2015b). This study takes one of the four underlying constructs of SeBIS
called Device Securement. It is a five-point Likert scale. For example, the item “I use a passcode
to unlock my mobile phone” is coded as 1 5 “Never”, 2 5 “Seldom”, 3 5 “Sometimes”, 407
4 5 “Often”, 5 5 “Always.” The total number of items on Device Securement scale is four. The
maximum score for Device Securement is 20 and the minimum score is 4.
3.1.3 Demographics and digital divide. In this study, gender is taken with 1 representing
males and 0 representing females. The age variable is taken as categorical values represent 2
groups: 18–20 years old coded as 1 were as individuals who are 21 and above years old are
coded as 0. Digital divide variable is taken as the frequency of Internet access. The individual
who access the Internet once a day was coded as 0, whereas those who had more access to
Internet were coded as 1 which represents individuals’ multiple times a day access to the
internet.
3.1.4 Validity and reliability of measures. The validity and reliability of the measures used
is an important factor in the survey-based studies. Validity is the extent to which a
questionnaire measures what is intended. The content validity assesses how well the
concepts of the scales are being translated into items. According to Straub (1989), expert
evaluations and reviews of the literature are ways to establish content validity. PISP and

Demographics Frequency Percentage

Gender
Male 159 56.0
Female 125 44.0
Age
18–20 161 56.7
21 and above 123 43.3
Frequency of Internet access
Once a day or less 38 13.4 Table 2.
Multiple times a day 246 86.6 Demographics

Self-disclosure Yes % No %

Do you have your profile online that is visible? 245 86.20 39 13.80
Do you allow anyone to see your profile? 201 70.77 83 29.23
Do you include picture in your social media account? 194 68.30 90 31.70
Do you include your email address in your social media account? 186 65.49 198 34.51
Do you include your instant messenger address in your social media account? 156 54.92 128 45.08
Do you include your phone number on your social media account? 132 46.40 152 53.60
Do you include your home address on your social media account? 89 31.30 195 68.70
Do you include your interests and hobbies in your social media account? 180 63.30 104 36.70
Do you write about your personality in your social media account? 174 61.20 110 38.80
Do you write comments about other people’s profile page? 158 55.60 126 44.40 Table 3.
Do you use your real name in your social media account? 255 89.70 29 10.30 Frequency analysis of
Do you spend time personalizing your social media account? 151 53.20 133 46.80 self-disclosure
K Device Securement scales have been content validated by an expert panel with the latter
52,1 making use of literature review as well. The criterion validity is the extent to which a scale
correlates with an expected variable/construct with which it is very likely to be correlated
(Bollen, 1989). The Device Securement scale has been validated by correlating it with existing
psychometrics (Egelman and Peer, 2015b). Its criterion validity has also been established due
to high correlations with relevant computer security behavior (Egelman et al., 2016). The
external validity is the generalization of the findings in other environments than that of the
408 study (Straub et al., 2004). This study incorporated six universities that were geographically
distributed. The universities have in their undergraduate programs’ respondents who are
active social media users. Therefore, the findings are generalizable and represent a true cross-
section of the population. The reliability is the extent to which the instrument measures
consistently. The Cronbach’s alpha was used for measuring reliability which should be
greater than or equal to 0.7 (Nunnally, 1994). The original values of Cronbach’s alpha for PISP
is 0.94 while that of device securement is 0.81 which shows high internal consistency.

3.2 Procedure
3.2.1 Data collection. The survey was administered via pen and paper strategy that took
10–15 min to fill. All the participants were informed about the anonymity of their responses.
The sample size adequacy was established a priori. In order for the results to be statistically
valid, the sample size should be at least 300 or 5 times greater than the total number of items
on a scale (Nunnally, 1994). Moreover, the response rate must be at least 30% to avoid non-
response bias (Nunnally, 1994). In this study, the calculated adequate sample size was 60;
therefore, the sample size adequacy was established.
3.2.2 Sampling and participants. The population considered for this study is the university
going students. A total of 300 students from 6 different universities filled in the questionnaire.
The universities were randomly selected from geographically distributed areas in the country
to ensure a diverse sample. After data cleaning, 284 responses were used for subsequent data
analysis. The response rate was 94.6% which was more than adequate to mitigate non-
response bias. The sample consisted of 56% males and 44% females and more than 50%
between the ages of 18 and 20 (Table 2).
Responses were analyzed using SPSS version 21. Descriptive statistics were used to
understand the data. Each variable was examined and reported with respect to
demographics. Independent samples t-test was used to find the differences in self-
disclosure of the participants with respect to gender, age and frequency of Internet access.
Multiple hierarchal regression was used to test the hypothesis. Moreover, AI-based predictive
analysis was carried out to augment the results of statistical regression using Python Lib
scikit-learn version 0.23. This study uses area under the curve–receiver operating
characteristic (AUC-ROC) measure to predict the accuracy of results. AUC-ROC is the
performance measurement that is used to measure the capability of the model in predicting
classes (Ling et al., 2003). Higher the value of AUC-ROC, the better is the model at prediction.
To see the details of the prediction and accuracy of classification of all the classes, a confusion
matrix has also been used hence the limitations of using AUC-ROC alone is overcome. The
confusion matrix is the overall summarized detail of prediction results in terms of total counts
belonging to a particular class in the classification model (Ling et al., 2003).

4. Results
4.1 Descriptive analysis
We did a frequency analysis of the self-disclosure habits that the users exhibit. Table 3
presents the summary in which it is ascertained that users self-disclose to send out
information about their interests such as 55% writing about their personality and 63% Hybrid
talking about their hobbies. The users were also engaged in disclosure of sensitive artificial
information with 46% sharing their phone numbers and 31% sharing their home addresses
on their social media profile. These percentages show that self-disclosure of information was
intelligence
observed in the sample ranging from simple information to highly sensitive one. This study approach
further investigates gender differences in terms of self-disclosure. The independent samples
t-test highlights the statistically significant differences between the male and female users in
their self-disclosure practices with (t 5 7.925, p < 0.01). The males (X5 8.48, SD 5 2.575) 409
were indulged in self-disclosure more than females (X5 6.16, SD 5 2.273). Similarly,
statistically significant differences were found between users who accessed the Internet more
and less frequently (t 5 3.446, p 5 0.001). Those who accessed the Internet more frequently
(X5 7.67, SD 5 2.62) disclosed information more than those who accessed the Internet less
frequently (X5 6.08, SD 5 2.84).

4.2 Predicting self-disclosure


A multiple hierarchal regression analysis was carried out to test the hypotheses of this study.
The results of correlation analysis (Table 4) show significant correlations among the
variables. The ordering of independent variables in multiple regression analysis can result in
spurious results (Field, 2009). The hierarchal regression can explain this variance by
portioning and ordering the independent variables. Therefore, we first used previously
known factors followed by cyber protection behavior in a stepwise manner. The analysis was
performed in three steps. In the first step, demographic variables – age and gender – were
entered. In the second step, the frequency of Internet access was added to the model (Figure 1).
In the third step, “cyber protection behavior” was entered into the model. The results of the
hierarchal regression (Table 5) showed that all blocks of predictors significantly explained

Variables 1 2 3 4 5

1. Self-disclosure 1
2. Age 0.68 1
3. Gender 0.427** 0.45 1
4. Frequency of Internet access 0.201** 0.051 0.006 1 Table 4.
5. Cyber protection behavior 0.196** 0.155** 0.022 0.248** 1 Correlations between
Note(s): **p < 0.01 the variables

Variable b β t R R2 Adj. R2 p-value

Step 1
Age 0.265 0.049 0.902 0.429 0.184 0.179 0.368
Gender 2.306 0.425 7.874 0.000
Step 2
Age 0.210 0.039 0.731 0.472 0.223 0.215 0.465
Gender 2.302 0.424 8.064 0.000
Frequency of Internet access 1.557 0.197 3.727 0.000
Step 3
Age 0.810 0.015 0.284 0.498 0.248 0.237 0.777
Gender 2.329 0.429 8.270 0.000 Table 5.
Frequency of Internet access 1.243 0.157 2.928 0.004 Results of multiple
Cyber protection behavior 0.108 0.165 3.037 0.003 hierarchal regression
K the variance of self-disclosure. Model 1 explained 18% of the variation in self-disclosure
52,1 (R2 5 0.184, F 5 31.787, df 5 2, p < 0.01). The introduction of frequency of Internet access
significantly explained the self-disclosure with the variation of 22% (R2 5 0.223, F 5 26.79,
df 5 3, p < 0.01). Similarly, with the addition of cyber protection behavior, Model 3 also
explained the self-disclosure significantly when controlling for demographic and frequency
of Internet access (R2 5 0.248, F 5 22.99, df 5 4, p < 0.01) with the variance of 24%.
The regression coefficients were examined for testing the four hypotheses. Regarding
410 demographics, H2 predicted that gender influences self-disclosure. Our regression results
confirmed the H2. Gender positively influences the self-disclosure on SNSs (β 5 0.425,
p < 0.01). The standardized effect of gender was lower when the frequency of Internet access
was added but was the highest with all the coefficients present in Model 3. Gender had the
highest β for all regressors. The third hypothesis (H3) predicted the influence of age on self-
disclosure negatively. The regression results did not find any significant association between
the two (β 5 0.039, p > 0.05). Hence, H3 was not supported. The fourth hypothesis (H4)
postulates the frequency of Internet access to have a positive influence on self-disclosure. As
hypothesized, the digital connectivity of the students was positively related to the self-
disclosure on SNSs (β 5 0.1975, p < 0.01). The standardized effects of frequency of Internet
access were lower in Model 3. Similarly, H1 was also supported by the addition of the cyber
protection behavior variable. The cyber protection behavior of the students positively
influenced online information sharing and predicted the self-disclosure significantly
(β 5 0.165, p 5 0.003). The cyber protection behavior of the students increased the
explanatory power of the model substantially (adj R2 5 0.237) with an increase of 10%.

4.3 Artificial intelligence-based predictive analysis


We tested our predictive model using the five classification algorithms listed in Table 1. This
demanded some pre-processing steps including the discretization of the cyber protection
behavior score into three classes, namely, low, medium and high and the self-disclosure into
two classes low and high according to the rules listed in Table 6. The variables were further
numeric encoded as 0 (low), 1 (medium) and 2 (high) for cyber protection behavior score and
0 (low) and 1(high) for self-disclosure. The classifiers were tested under two settings: (1) 20%
hold-out (HO) method which used 80% of the data for training and the remaining 20% to
evaluate the performance of the classifier. (2) A 10-fold cross-validation (CV) testing was also
performed to mitigate the effects of overfitting for more accurate performance scores (Bzdok
et al., 2018).
Table 7 presents the accuracy, AUC-ROC and F1 scores for each of the five models under
the HO and 10-fold CV settings and the fitting times under CV training. The CV scores
(Table 7) are consistently lower than HO values for each metric which is indicative of
overfitting in the HO setting (Bzdok et al., 2018). Hence, we will consider the 10-fold CV scores
for the rest of the discussion. Moreover, literature establishes the superiority of AUC-ROC
scores for classifier performance assessment over accuracy when the training set is

Variables Min Max Mean SD Low Medium High

Gender 0.0 1.0 – – – – –


Age 0.0 1.0 – – – – –
Frequency of internet access 0.0 1.0 0.86 0.34 – – –
Table 6. Cyber protection behavior 4.0 20.0 16.18 4.09 <(μ – 1 SD) (μ – 1 SD) - μ >μ
Attributes <12.08 12.08–16.179 >16.179
characteristics for ML Self-disclosure 0.0 12.0 7.45 2.7 <5μ - >μ
predictive analysis <5 7.4577 > 7.4577
imbalanced (Jin Huang and Ling, 2005). Thus, we will consider the AUC-ROC and F1-scores to Hybrid
compare model performances. Figure 2 presents the area under the receiver operating artificial
characteristics (ROC) curve also known as the area under the curve (AUC). It shows the
classifiers on test data under HO testing. Naı€ve Bayes shows the highest AUC-ROC and F1-
intelligence
measure of 0.744 and 0.7010, respectively. KNN is the lowest scorer with 0.594 AUC-ROC and approach
0.624 F1-measure. This is consistent with the fact that KNN works better with continuous
numeric features for which distance metrics are more meaningful. Figures 3–7 show the
411

Accuracy AUC-ROC F1-measure


Algorithms 10-Fold CV TTS 10-Fold CV TTS 10-Fold CV TTS Training time

LR 0.69347 0.75438 0.73824 0.75369 0.70311 0.7586 0.0797


NB 0.71477 0.77192 0.74411 0.76477 0.70953 0.76363 0.0349
DT 0.69716 0.71929 0.68057 0.728448 0.682507 0.71428 0.0309
SVM 0.67931 0.701754 0.7344 0.80172 0.67721 0.72131 0.1291 Table 7.
KNN 0.63374 0.66666 0.59368 0.69841 0.62410 0.73460 0.0286 ML algorithms
Note(s): The italics emphasize the highest AUC-ROC and F1 measures of NB algorithm performance analysis

Figure 2.
The AUC-ROC of ML
algorithms
K
52,1

412

Figure 3.
Logistic
Regression (LR)

Figure 4.
Naı€ve Bayes (NB)

confusion matrices presenting the summary of prediction results by each classifier as counts
of correct and incorrect classifications against each class. The classification results confirm
that age, gender, frequency of Internet access and cyber protection behavior successfully
predicted the level of self-disclosure with sufficient accuracy which further endorses our
earlier findings from statistical inference.

5. Discussion
Self-disclosure on SNSs has opened up many opportunities for individuals to indulge in
capital-enhancing activities (Tzortzaki et al., 2016). It is associated with subjective well-being,
improved connectedness, relationship gains, social capital and relief from loneliness.
Individuals who avidly disclose information about themselves by posting pictures and
comments invariably gain more attention from their friends, peers and family members. On
Hybrid
artificial
intelligence
approach

413

Figure 5.
Decision Tree (DT)

Figure 6.
Support vector
machine (SVM)

the other hand, self-disclosure also leads to privacy violations and cybercrimes such as
identity theft, blackmail and kidnapping. Therefore, individuals make use of security and
privacy safeguards to maintain the benefits of self-disclosure while mitigating the threats
arising from the disclosure of information. This study provides insights into cybersecurity
awareness particularly protective behavior to influence self-disclosure by the enactment of
cost-benefit analysis.
The study hypothesized that the cyber protection behavior of the users will be
significantly associated with self-disclosure and found a positive association between the
two. The model was based on the EVT and hypotheses were tested by using a multiple
hierarchal approaches. The results show that the expectancy of taking cybersecurity
measures enhances individual’s self-disclosure behavior thus explaining the cost-benefits
consideration. This indicates that the more strictly the users exhibit cyber protection
behavior, the more they are self-confident in their ability to protect themselves which in turn
K
52,1

414

Figure 7.
K-nearest neighbor
(KNN) K 5 5

will result in maximization of their self-disclosure. These findings are in line with a similar
study (Rauhala et al., 2019), in which spending on cybersecurity to protect themselves online
positively influences online expression. Other studies have also found that cybersecurity
concepts such as anonymity – which is the degree to which the identity is kept under the wrap
while being online – has also been associated with increased self-disclosure. Individuals who
maintain themselves anonymous via the use of pseudonyms exhibit increased levels of
disclosure of personal information on SNSs (Clark-Gordon et al., 2019). This cyber protection
behavior is a prominent antecedent of self-disclosure phenomenon as reported in the
literature thereby corroborating the findings of this study.
Individuals especially youngsters take security precautions to reap the benefits of self-
disclosure. The younger individuals who are enrolled in tertiary institutes are avid users of
SNSs and share information including photos, ideas and status updates to present themselves
in their social networks. At the same time, they are the most vulnerable group to cybercrimes
(Kortjan et al., 2012). In order to maintain their presence in their social networks while taking
precautions against cybercriminals, they make use of security settings. Literature shows that
these individuals are concerned about their security (Pramod and Raman, 2014). Moreover,
the university going students possess knowledge about cybersecurity practices (Kim, 2014);
therefore, the valence of being technically well advanced determines their choice to self-
disclose on SNSs. These young individuals have a better know-how of applying security
settings than any other group hence have self-confidence in practicing cyber protection
behavior to gain benefits associated with disclosure of information. Our findings are also in
line with studies that predict self-disclosure behavior with respect to protection behaviors
exhibited by the use of technical settings in different online platforms (Arpaci, 2020; Stern and
Salb, 2015; Young and Quan-Haase, 2013).
Our study supports hypothesis related to the influence of cyber protection behavior on
self-disclosure as well as the other two hypotheses related to gender and frequency of Internet
access. The results of this study are in line with previous studies regarding gender
(Almakrami, 2015; Al-Saggaf, 2016; Bond, 2009; Hoy and Milne, 2010; Sheldon, 2013; Wang
et al., 2016) in which gender has an influence on self-disclosure. On the other hand, the results
are inconsistent with previous studies regarding age (Malhotra et al., 2004) having no
influence on self-disclosure. The influence of frequency of Internet access on self-disclosure in
this study corroborates the findings of Shin and Kang (2016), Xie and Kang (2015) and
Błachnio et al. (2016). Younger individuals are an avid user of the SNSs; however, the Hybrid
frequency of the use of Internet varies depending upon digital facilities available to them. artificial
Past studies have reported that almost 25% of the younger individuals are constantly using
the Internet (Koohikamali et al., 2017). The frequency of Internet access has also been
intelligence
associated with other online behaviors. In fact, the digital divide affects the digital skills and approach
knowledge of individuals which indirectly influence their online behavior (Dodel and Mesch,
2019). This suggests that individuals who are digitally less connected may miss out on the
social capital and relationship gains due to limited self-disclosure and at the same time have 415
lower cyber protection behavior. The statistical results of this study are augmented with the
AI-based predictive analysis. ML algorithms have been used in other studies to augment the
statistical results in the study of human behaviors (Arpaci, 2019, 2020). Moreover, these
techniques have been able to detect such behavior with an accuracy of 80–90%
(Wiederhold, 2019b).

6. Contribution and practical implications


This study contributes toward theory and practice to augment already present literature in
explaining self-disclosure. A previous systematic literature review (Abramova et al., 2017)
has reported the effects of technical privacy measures on self-disclosure. This study extends
the knowledge in the field by providing evidence that cybersecurity awareness influences
self-disclosure in the light of EVT. Previous studies fall short on the application of theoretical
models to understand the influence of cybersecurity awareness on self-disclosure. Moreover,
the cybersecurity awareness is investigated in terms of users’ protection behavior rather than
their spending on security software which may not capture the actual usage of security
software.
The results of this study have important practical implications in terms of understanding
the trade-off between security on SNSs and perceived benefits. Practically, this work is
beneficial to cybersecurity practitioners and researchers in understanding the influence of
cybersecurity awareness on different information-related behaviors. In cybersecurity
research, self-disclosure is associated with negative consequences of Internet experience.
High disclosures lead to potential cybercrimes and hence are considered relatively unsafe.
This is particularly true for the younger generation specifically the university going students.
As per Albert and Steinberg (2011), the propensity of the youngster to engage in online self-
disclosure is comparatively more than any other age group. These young individuals share
their day-to-day activities on social media to present themselves on their social networks.
This cyberbehavior of continuously sharing personal information is now considered a norm
and is a priority for youngster (Aizenkot, 2020). The constant information feed in these
platforms can be used to enact cyber-profiles of these youngsters which can then be beneficial
for criminals to carry out different crimes. The intensity of these cybercrimes varies from
being low to very high. Anecdotal evidence shows that the information gleaned from social
networks can be used to incur financial losses and character assassination by editing photos
and posting on pornographic sites. On the darker side, kidnappings and even loss of life have
been attributed toward self-disclosure. However, despite the dangers, the amount of self-
disclosure cannot be controlled; neither can one expect to lower it due to the benefits that are
related to it. As shown by existing studies, the use of SNSs is on the rise despite adverse
effects on individual’s well-being (Dhir et al., 2021). This means that the youngsters will
continue to self-disclose attributing toward the benefits of self-presentation, relationship
gains, entertainment and the prospects of deriving social and monetary capital. As has been
reported by Barth et al. (2019), the benefits of low costs and functionality while downloading a
mobile app have been found to outweigh the privacy risks. Therefore, enforcing social media
limitations is not an optimal solution as it will limit the benefitting activities. Instead of
K limiting the capital-enhancing activities of the Internet, university students should be trained
52,1 to become technically proficient in cyber protection practices. The educational interventions
used to tackle the negative consequences of self-disclosure should be focused on the capacity
building of the users. Users, when they are proficient in cybersecurity technical measures on
their devices will feel confident and enjoy the capital-enhancing benefits of self-disclosure.
Moreover, these trainings should be tailored keeping in mind those students who are digitally
less connected since they possess lesser digital knowledge and skills compared to those who
416 have constant Internet access. These specialized trainings should incorporate the benefits of
self-disclosure as well as the risk associated with it. A small portion of specific training should
also touch upon the implications of cyber protection behaviors in a safe and secure experience
of the Internet.
This study also contributes to inform tertiary institutes’ management. The unrestricted
privacy risk associated with sharing of information and the openness in SNSs is the cause of
concern for the university management. On the other hand, the open environment and
exchange of information in SNSs allow the students to discuss ideas, collaborate and interact
among themselves and gain benefits. With scientific evidence promoting control of SNSs and
formulation and execution of policies in corporate organizations (Soomro et al., 2016), the
same cannot be adopted in academic organizations. The extended use of SNSs for information
sharing involving academic activities will take a substantiated hit if policies are enacted with
a ban on or control of these sites. Rather effective and efficient policies are needed which allow
for reaping the benefits of SNSs and consequential self-disclosure while at the same time
ensuring the privacy and security of the students. The security awareness programs should
be incorporated into the tertiary institutes’ policies that are to be delivered consistently with a
pre-set time frame. The security landscape is volatile and is constantly changing due to the
advancement in technology. This changing landscape requires a regular update in the
cybersecurity awareness programs that should be delivered periodically to improve and
reinforce cyber protection practices. Moreover, governments should take into consideration
that these students are the future workforce and will be participants in the economic
development of the country. In order for these individuals to avoid being victims of the
cybercrimes and incurring any corporate losses due to potential risky self-disclosure in the
organizational settings, governments should be a stakeholder in the development, execution
and maintenance of such awareness programs.

7. Limitations and future directions


The sample for this work was drawn from the student population. The younger users are
more adaptive to the technology and spend most of their time on SNSs; therefore, we believe
that they are a representative sample in the context of this study. Another limitation of the
study is the reliance on self-report data which may inculcate measurement errors and biases.
The dispositional and situational biases were taken care of by assuring the anonymity of the
responses. Future work may include understating other factors that have already been
shown to have an impact on the disclosure of information such as trust and self-efficacy with
cybersecurity awareness in a cross-cultural context.

8. Conclusion
In conclusion, this study is among the first to investigate the influence of cyber protection
behavior on self-disclosure using a hybrid regression and AI-based approach. The results are
positive and contribute toward the self-disclosure body of knowledge by providing key
insights to researchers which are beneficial for practitioners as well. This study provides
starting evidence that the cybersecurity awareness’s role is significant in understanding
self-disclosure. The benefits of self-disclosure are enjoyed by the users and the decision is Hybrid
taken in a calculated way by their proficiency in the security settings. Users are willing to self- artificial
disclose to enjoy the benefits associated with it and downplay the risks and dangers by their
ability to protect themselves.
intelligence
approach
References
Abramova, O., Wagner, A., Krasnova, H. and Buxmann, P. (2017), “Understanding self-disclosure on 417
social networking sites - a literature review”, The 23rd Americas Conference on Information
Systems (AMCIS), Boston.
Aharony, N. (2016), “Relationships among attachment theory, social capital perspective, personality
characteristics, and Facebook self-disclosure”, Aslib Journal of Information Management,
Vol. 68 No. 3, pp. 362-386.
Aivazpour, Z. and Rao, V.S. (2020), “Information disclosure and privacy paradox: the role of
impulsivity”, ACM SIGMIS Database: The Database for Advances in Information Systems,
Vol. 51, pp. 14-36.
Aizenkot, D. (2020), “Social networking and online self-disclosure as predictors of cyberbullying
victimization among children and youth”, Children and Youth Services Review, Vol. 119, 105695.
Ajis, A. and Salleh, N.A.M. (2020), “Core elements of motivational factors that influence Facebook
users to self-disclosure”, International Journal of Business Management, Vol. 4, pp. 1-11.
Al-Saggaf, Y. (2016), “An exploratory study of attitudes towards privacy in social media and the
threat of blackmail: the views of a group of Saudi women”, The Electronic Journal of
Information Systems in Developing Countries, Vol. 75.
Albert, D. and Steinberg, L. (2011), “Judgment and decision making in adolescence”, Journal of
Research on Adolescence, Vol. 21, pp. 211-224.
Almakrami, H.A.A. (2015), “Online self-disclosure across cultures: a study of Facebook use in Saudi
Arabia and Australia”, PhD thesis, Queensland University of Technology.
Alshare, K.A., Moqbel, M. and Al-Garni, M.A. (2019), “The impact of trust, security, and privacy on
individual’s use of the internet for online shopping and social media: a multi-cultural study”,
International Journal of Mobile Communications, Vol. 17, pp. 513-536.
Ampong, G.O.A., Mensah, A., Adu, A.S.Y., Addae, J.A., Omoregie, O.K. and Ofori, K.S. (2018),
“Examining self-disclosure on social networking sites: a flow theory and privacy perspective”,
Behavioral Sciences, Vol. 8, p. 58.
Anderson, K.B., Durbin, E. and Salinger, M.A. (2008), “Identity theft”, Journal of Economic
Perspectives, Vol. 22, pp. 171-192.
Arpaci, I. (2019), “A hybrid modeling approach for predicting the educational use of mobile cloud
computing services in higher education”, Computers in Human Behavior, Vol. 90, pp. 181-187.
Arpaci, I. (2020), “What drives students’ online self-disclosure behaviour on social media? A hybrid
SEM and artificial intelligence approach”, International Journal of Mobile Communications,
Vol. 18, pp. 229-241.
B Kim, E. (2014), “Recommendations for information security awareness training for college students”,
Information Management and Computer Security, Vol. 22, pp. 115-126.
Błachnio, A., Przepiorka, A., Bałakier, E. and Boruch, W. (2016), “Who discloses the most on
Facebook?”, Computers in Human Behavior, Vol. 55, pp. 664-667.
Ball, A.L., Ramim, M.M. and Levy, Y. (2015), “Examining users’ personal information sharing
awareness, habits, and practices in social networking sites and e-learning systems”, Online
Journal of Applied Knowledge Management, Vol. 3, pp. 180-207.
Barrett-Maitland, N., Barclay, C. and Osei-Bryson, K.-M. (2016), “Security in social networking
services: a value-focused thinking exploration in understanding users’ privacy and security
concerns”, Information Technology for Development, Vol. 22, pp. 464-486.
K Barth, S. and De Jong, M.D. (2017), “The privacy paradox–Investigating discrepancies between
expressed privacy concerns and actual online behavior – a systematic literature review”,
52,1 Telematics and Informatics, Vol. 34, pp. 1038-1058.
Barth, S., de Jong, M.D., Junger, M., Hartel, P.H. and Roppelt, J.C. (2019), “Putting the privacy paradox
to the test: online privacy and security behaviors among users with technical knowledge,
privacy awareness, and financial resources”, Telematics and Informatics, Vol. 41, pp. 55-69.
Biran, O. and McKeown, K.R. (2017), “Human-centric justification of machine learning predictions”,
418 IJCAI, pp. 1461-1467.
Bollen, K.A. (1989), “A new incremental fit index for general structural equation models”, Sociological
Methods and Research, Vol. 17, pp. 303-316.
Bond, B.J. (2009), “HE posted, she posted: gender differences IN self-disclosure ON social network
sites”, Rocky Mountain Communication Review, Vol. 6.
by, C. and Gaggioli, A. (2017), “Bringing more transparency to artificial intelligence”, Cyberpsychology,
Behavior, and Social Networking, Vol. 20, p. 68.
Bzdok, D., Altman, N. and Krzywinski, M. (2018), “Statistics versus machine learning”, Nature
Methods, Vol. 15 No. 4, pp. 233-234, doi: 10.1038/nmeth.4642.
Chennamaneni, A. and Taneja, A. (2015), “Communication privacy management and self-disclosure on
social media - a case of Facebook”, The 2015 Americas Conference on Information Systems
(AMCIS), Puerto Rico.
Cheung, C., Lee, Z.W. and Chan, T.K. (2015), “Self-disclosure in social networking sites”, Internet
Research, Vol. 25 No. 2, pp. 279-299.
Clark-Gordon, C.V., Bowman, N.D., Goodboy, A.K. and Wright, A. (2019), “Anonymity and online self-
disclosure: a meta-analysis”, Communication Reports, Vol. 32, pp. 98-111.
Dhir, A., Talwar, S., Kaur, P., Budhiraja, S. and Islam, N. (2021), “The dark side of social media:
stalking, online self-disclosure and problematic sleep”, International Journal of Consumer
Studies, Vol. 45 No. 6, doi: 10.1111/ijcs.12659.
Dodel, M. and Mesch, G. (2019), “An integrated model for assessing cyber-safety behaviors: how
cognitive, socioeconomic and digital determinants affect diverse safety practices”, Computers
and Security, Vol. 86, pp. 75-91.
Egelman, S. and Peer, E. (2015a), “Predicting privacy and security attitudes”, ACM SIGCAS
Computers and Society, Vol. 45, pp. 22-28.
Egelman, S. and Peer, E. (2015b), “Scaling the security wall: developing a security behavior intentions
scale (sebis)”, in Proceedings of the 33rd Annual ACM Conference on Human Factors in
Computing Systems, ACM, pp. 2873-2882.
Egelman, S., Harbach, M. and Peer, E. (2016), “Behavior ever follows intention? A validation of the
security behavior intentions scale (SeBIS)”, in Proceedings of the 2016 CHI Conference on
Human Factors in Computing Systems, pp. 5257-5261.
Facebook, Twitter, options, S. more sharing, Facebook, Twitter, LinkedIn, Email, URLCopied!, C.L.,
Print (2020), “Men used social media to contact teen who was kidnapped and sexually
assaulted, police say”, [WWW Document], Los Angeles Times, available at: https://www.latimes.
com/california/story/2020-01-07/men-used-social-media-to-contact-teen-who-was-kidnapped-
and-sexually-assaulted-police-say (accessed 7 October 2021).
Field, A. (2009), Discovering Statistics Using SPSS, Sage Publications, London.
Fogel, J. and Nehmad, E. (2009), “Internet social network communities: risk taking, trust, and privacy
concerns”, Computers in Human Behavior, Vol. 25, pp. 153-160.
Furnell, S. (2008), “End-user security culture: a lesson that will never be learnt?”, Computer Fraud and
Security, pp. 6-9.
Gaggioli, A. (2017), “Artificial intelligence: the future of cybertherapy?”, Cyberpsychology, Behavior,
and Social Networking, Vol. 20, pp. 402-403.
Gahagan, K., Vaterlaus, J.M. and Frost, L.R. (2016), “College student cyberbullying on social Hybrid
networking sites: conceptualization, prevalence, and perceived bystander responsibility”,
Computers in Human Behavior, Vol. 55, pp. 1097-1105. artificial
Gharibi, W. and Shaabi, M. (2012), Cyber Threats in Social Networking Websites, arXiv preprint arXiv:
intelligence
1202.2420. approach
Grieve, R., Indian, M., Witteveen, K., Tolan, G.A. and Marrington, J. (2013), “Face-to-face or Facebook:
can social connectedness be derived online?”, Computers in Human Behavior, Vol. 29,
pp. 604-609. 419
Gupta, A. and Dhami, A. (2015), “Measuring the impact of security, trust and privacy in information
sharing: a study on social networking sites”, Journal of Direct, Data and Digital Marketing
Practice, Vol. 17, pp. 43-53.
Hajli, N. and Lin, X. (2016), “Exploring the security of information sharing on social networking sites:
the role of perceived control of information”, Journal of Business Ethics, Vol. 133, pp. 111-123.
Hallam, C. and Zanella, G. (2017), “Online self-disclosure: the privacy paradox explained as a
temporally discounted balance between concerns and rewards”, Computers in Human Behavior,
Vol. 68, pp. 217-227.
Haskell-Dowland, P. (n.d), “Facebook data breach: what happened and why it’s hard to know if your
data was leaked”, [WWW Document], The Conversation, available at: http://theconversation.
com/facebook-data-breach-what-happened-and-why-its-hard-to-know-if-your-data-was-leaked-
158417 (accessed 7 9 21).
Heravi, A., Mubarak, S. and Choo, K.-K.R. (2018), “Information privacy in online social networks: uses
and gratification perspective”, Computers in Human Behavior, Vol. 84, pp. 441-459.
Hoy, M.G. and Milne, G. (2010), “Gender differences in privacy-related measures for young adult
Facebook users”, Journal of Interactive Advertising, Vol. 10, pp. 28-45.
Huang, J. and Ling, C.X. (2005), “Using AUC and accuracy in evaluating learning algorithms”, IEEE
Transactions on Knowledge and Data Engineering, Vol. 17, pp. 299-310, doi: 10.1109/TKDE.
2005.50.
Kefi, H. and Perez, C. (2018), “Dark side of online social networks: technical, managerial, and
behavioral perspectives”, in Encyclopedia of Social Network Analysis and Mining, pp. 1-22. doi:
10.1007/978-1-4614-7163-9_110217-1.
Koohikamali, M., Peak, D.A. and Prybutok, V.R. (2017), “Beyond self-disclosure: disclosure of
information about others in social network sites”, Computers in Human Behavior, Vol. 69,
pp. 29-42.
Kortjan, N., von Solms, R. and Van Niekerk, J. (2012), “Ethical guidelines for cyber-related services
aimed at the younger generations”, Proceedings of the Sixth International Symposium on
Human Aspects of Information Security & Assurance, Crete, pp. 205-215.
Li, P., Cho, H. and Goh, Z.H. (2019), “Unpacking the process of privacy management and self-
disclosure from the perspectives of regulatory focus and privacy calculus”, Telematics and
Informatics, Vol. 41, pp. 114-125.
Ling, C.X., Huang, J. and Zhang, H. (2003), “AUC: a better measure than accuracy in comparing
learning algorithms”, in Conference of the Canadian Society for Computational Studies of
Intelligence, Springer, pp. 329-341.
LinkedIn Data Breach - 500M Records Leaked and Being Sold [WWW Document] (2021), CyberNews,
available at: https://cybernews.com/news/stolen-data-of-500-million-linkedin-users-being-sold-
online-2-million-leaked-as-proof-2/ (accessed 7 September 21).
Loiacono, E.T. (2015), “Self-disclosure behavior on social networking web sites”, International Journal
of Electronic Commerce, Vol. 19, pp. 66-94.
Malhotra, N.K., Kim, S.S. and Agarwal, J. (2004), “Internet users’ information privacy concerns (IUIPC):
the construct, the scale, and a causal model”, Information Systems Research, Vol. 15,
pp. 336-355.
K Mouton, F., Leenen, L. and Venter, H.S. (2016), “Social engineering attack examples, templates and
scenarios”, Computers and Security, Vol. 59, pp. 186-209.
52,1
Mubarak, S. and Rahamathulla, M.A. (2016), Online Self-Disclosure and Wellbeing of Adolescents: A
Systematic Literature Review, arXiv preprint arXiv:1606.03527.
Nunnally, J.C. (1994), Psychometric Theory 3E, Tata McGraw-Hill Education, New Delhi.
Parsons, K., Calic, D. and Barca, C. (2016), “Self-disclosure on Facebook: comparing two research
420 organisations”, The Australasian Conference on Information Systems 2016, Wollongong.
Pramod, D. and Raman, R. (2014), “A study on the user perception and awareness of smartphone
security”, International Journal of Applied Engineering Research, Vol. 9 No. 23, pp. 19133-19144.
Pujazon-Zazik, M. and Park, M.J. (2010), “To tweet, or not to tweet: gender differences and potential
positive and negative health outcomes of adolescents’ social internet use”, American Journal of
Men’s Health, Vol. 4, pp. 77-85.
Rauhala, J., Tyrv€ainen, P. and Zaidenberg, N. (2019), “Online expression and spending on personal
cybersecurity”, in ECCWS 2019 18th European Conference on Cyber Warfare and Security,
Academic Conferences and Publishing, p. 387.
Robinson, S.C. (2017), “Self-disclosure and managing privacy: implications for interpersonal and
online communication for consumers and marketers”, Journal of Internet Commerce, Vol. 16,
pp. 385-404.
Salo, J., M€antym€aki, M. and Islam, A.N. (2018), “The dark side of social media–and Fifty Shades of
Grey introduction to the special issue: the dark side of social media”, Internet Research, Vol. 28
No. 5, pp. 1166-1168.
Segura, J. (2017), Compromised LinkedIn Accounts Used to Send Phishing Links via Private Message
and InMail, [WWW Document], Malwarebytes Labs, available at: https://blog.malwarebytes.
com/threat-analysis/2017/09/compromised-linkedin-accounts-used-to-send-phishing-links-via-
private-message-and-inmail/ (accessed 7 September 2021).
Shaw, R.S., Chen, C.C., Harris, A.L. and Huang, H.-J. (2009), “The impact of information richness on
information security awareness training effectiveness”, Computers and Education, Vol. 52,
pp. 92-100.
Sheldon, P. (2013), “Examining gender differences in self-disclosure on Facebook versus face-to-face”,
The Journal of Social Media in Society, Vol. 2.
Shin, W. and Kang, H. (2016), “Adolescents’ privacy concerns and information disclosure online: the
role of parents and the Internet”, Computers in Human Behavior, Vol. 54, pp. 114-123.
Shin, W., Huh, J. and Faber, R.J. (2012), “Developmental antecedents to children’s responses to online
advertising”, International Journal of Advertising, Vol. 31, pp. 719-740.
Singh, K.P. and Gill, M.S. (2015), “Role and users’ approach to social networking sites (SNSs): a study
of universities of North India”, The Electronic Library, Vol. 33 No. 1, pp. 19-34.
Soomro, Z.A., Shah, M.H. and Ahmed, J. (2016), “Information security management needs more holistic
approach: a literature review”, International Journal of Information Management, Vol. 36,
pp. 215-225, doi: 10.1016/j.ijinfomgt.2015.11.009.
Steijn, W.M. (2014), “A developmental perspective regarding the behaviour of adolescents, young
adults, and adults on social network sites”, Cyberpsychology: Journal of Psychosocial Research on
Cyberspace, Vol. 8.
Stern, T. and Salb, D. (2015), “Examining online social network use and its effect on the use of privacy
settings and profile disclosure”, Bulletin of Science, Technology and Society, Vol. 35, pp. 25-34.
Straub, D.W. (1989), “Validating instruments in MIS research”, MIS Quarterly, pp. 147-169.
Straub, D., Rai, A. and Klein, R. (2004), “Measuring firm performance at the network level: a nomology
of the business impact of digital supply networks”, Journal of Management Information
Systems, Vol. 21, pp. 83-114.
Stutzman, F., Capra, R. and Thompson, J. (2011), “Factors mediating disclosure in social network Hybrid
sites”, Computers in Human Behavior, Vol. 27, pp. 590-598.
artificial
Trepte, S. and Reinecke, L. (2013), “The reciprocal effects of social network site use and the
disposition for self-disclosure: a longitudinal study”, Computers in Human Behavior, Vol. 29,
intelligence
pp. 1102-1112. approach
Tzortzaki, E., Kitsiou, A., Sideri, M. and Gritzalis, S. (2016), “Self-disclosure, Privacy concerns and
Social Capital benefits interaction in FB: a case study”, in Proceedings of the 20th Pan-Hellenic
Conference on Informatics, pp. 1-6. 421
Van Deursen, A.J., Helsper, E., Eynon, R. and Van Dijk, J.A. (2017), “The compoundness and
sequentiality of digital inequality”, International Journal of Communication, Vol. 11,
pp. 452-473.
Van Dijk, J.A. (2005), The Deepening Divide: Inequality in the Information Society, Sage Publications,
University of Twente, Nethelands.
van Ingen, E. and Matzat, U. (2018), “Inequality in mobilizing online help after a negative life event:
the role of education, digital skills, and capital-enhancing Internet use”, Information,
Communication and Society, Vol. 21, pp. 481-498.
Vroom, V.H. (1964), Work and Motivation, Wiley, New York.
Walsh, R.M., Forest, A.L. and Orehek, E. (2020), “Self-disclosure on social media: the role of perceived
network responsiveness”, Computers in Human Behavior, Vol. 104, p. 106162.
Wang, T., Duong, T.D. and Chen, C.C. (2016), “Intention to disclose personal information via mobile
applications: a privacy calculus perspective”, International Journal of Information Management,
Vol. 36, pp. 531-542.
Webcam blackmail cases have doubled, police say (2016), BBC News.
Wheeless, L.R. and Grotz, J. (1976), “Conceptualization and measurement of reported self-disclosure”,
Human Communication Research, Vol. 2, pp. 338-346.
Wiederhold, B.K. (2019a), “Can artificial intelligence predict the end of life. . . and do we really want to
know?”, Cyberpsychology Behavior and Social Networking, Mary Ann Liebert, Vol. 22 No. 5,
pp. 297-299.
Wiederhold, B.K. (2019b), “Artificial intelligence and suicide: where artificial intelligence stops and
humans join in”, Cyberpsychology Behavior and Social Networking, Mary Ann Liebert, Vol. 22
No. 6, pp. 363-364.
Wilson, D. and Valacich, J.S. (2012), “Unpacking the privacy paradox: irrational decision-making
within the privacy calculus”, Proceedings of the 33rd International Conference on Information
Systems (ICIS2012), Florida.
Xie, W. and Kang, C. (2015), “See you, see me: teenagers’ self-disclosure and regret of posting on social
network site”, Computers in Human Behavior, Vol. 52, pp. 398-407.
Young, A.L. and Quan-Haase, A. (2013), “Privacy protection strategies on Facebook: the Internet
privacy paradox revisited”, Information, Communication and Society, Vol. 16, pp. 479-500.

Corresponding author
Naurin Farooq Khan can be contacted at: naurin.zamir@riphah.edu.pk

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: permissions@emeraldinsight.com

You might also like