Factors Affecting Personal Information Sharing: Small-Scale Sample Analysis on Social Media

— In the contemporary social landscape, the widespread use of social media, such as platforms like TikTok, Instagram, and YouTube, has become a prominent trend in various circles of society, especially in Indonesia. As the number of users on these platforms increases, concerns regarding user security and privacy also increase. Data breaches in 2021 affecting 235 million users on Instagram, TikTok, and YouTube underscored the importance of researching the multifaceted dynamics around privacy concerns, levels of trust, risk awareness, and user behavior patterns related to sharing personal information on social media platforms. This research aimed to address this critical issue by introducing a research model developed based on relevant hypotheses from previous research. The sample used in this research consisted of social media users in Indonesia. Methodologically, this research used sophisticated structural equation modeling (SEM) tools for hypothesis testing and confirmatory factor analysis (CFA) to validate the efficacy of existing research models. These findings indicated that users’ trust, awareness, privacy concerns, and behavioral intentions significantly and positively influence the tendency to share personal data on social media platforms. This research provides valuable insights into the complex interactions between factors influencing user behavior in social media privacy, thereby offering implications for academia and practical applications


I. INTRODUCTION
Social media is an Internet platform where people can easily interact, as well as exchange and produce content.Social media is something that is familiar to the public.Social media users are multiplying each year [1].They are not limited to age and social interaction on social media.In addition, they are not just limited to teenagers seeking popularity [2].This platform encourages individuals from diverse demographics of any kind by giving them access to open media to enhance the communication experience as well as to market and advertise their products [2].
According to a survey from the Indonesian Internet Service Providers Association (Asosiasi Penyelenggara Jasa Internet Indonesia, APJII), in the years 2022 to 2023, there were 215.63 million Internet users in Indonesia.This number climbed by 2.67% compared to the prior time frame, which had 210.03 million users.Among the total of 215.63 internet users, 89.15% are social media users [3].
One social media platform widely used by Indonesians is YouTube.YouTube ranks second with 2,514 million active users, Instagram ranks third with 2,000 million active users, and TikTok ranks sixth with 1,051 million [1].Although YouTube claims second place and Instagram takes third place, it is essential to note that the popularity and user engagement of social media platforms may vary based on regions, demographics, and other factors.
There are several strong reasons for the decision to focus this research on Instagram, TikTok, and YouTube.First, these three platforms represent a wide range of formats and user interactions, including photo sharing (Instagram), short-form video content (TikTok), and longer-form video content (YouTube).It allows for a comprehensive exploration of user behavior and attitudes across various types of media consumption.
Second, Instagram, TikTok, and YouTube have different user demographics and content dynamics, making them interesting subjects for comparative analysis.Instagram, for example, is known for its visual-centric content and influencer culture; TikTok thrives on creative and viral short-form videos; meanwhile, YouTube caters to a variety of content creators producing longer-form videos.Studying these platforms individually and collectively provides a nuanced understanding of user behavior, privacy concerns, and trust dynamics.
Finally, the decision to focus on Instagram, TikTok, and YouTube were influenced by the prevalence and impact of these platforms on contemporary social interactions and information-sharing practices.Instagram, TikTok, and YouTube have emerged as influential platforms shaping digital culture, and understanding user perceptions and behavior on these platforms provides valuable insight into broader social media trends.
The more developed social media is, the more important the issue of information security and privacy.Nowadays, social media, as one of the sources of leaking confidential information, has become a common thing.Without realizing it, a lot of personal data about someone has been leaked on the Internet.In Indonesia itself, the most recent case of personal data leakage is the personal leak of Instagram, TikTok, and YouTube users.Recently, hundreds of these users' data have been shared by an amateur hacker forum user for free.Some of the leaked data belong to celebrities, including food bloggers, celebrity figures, and other influencers [4].Data leaks have also been experienced by one of the Indonesian influencers' social media accounts, which then impact their trust in social media safety and reputation [5].
Many privacy issues in social media, such as trust, awareness, security, and behavioral intentions of users when using social media, need to be addressed.Unscrupulous individuals can illegally use users' data, including multimedia data for financial benefit.Risks associated with personal data shared on users' profiles include online harassment, online stalking, and identity theft.In addition, the main feature of social media itself is one of the supporters of data leakage.These features including tagging, sharing images, and comments, can increase leakage of users' personal information and make it more accessible to unauthorized parties [6], [7].
The existing social media privacy-related research has focused only on individual factors such as trust [8]; risk [9]; controls [10], [11]; as well as privacy concerns and user awareness [12].This research is based on the findings of previous studies, especially [4], [12], [13].The main aim of this research is to investigate the influence of users' understanding, confidence, and concerns regarding confidentiality on their behavior when sharing data on social networking sites.Using the structural equation modeling (SEM) method, this research explores the relationships between variables.The results showed that users' concerns about confidentiality, security, and trust on social media, coupled with their understanding of data security, had a significant and positive impact on their willingness to disclose personal data on those platforms.Additionally, key findings suggested that users' level of trust in social media services' ability to protect their personal information was also influenced by concerns about privacy and user understanding.In addition, increasing user privacy awareness has the potential to motivate users to increase privacy security on their social media accounts.Therefore, in order to differentiates it from previous research efforts, this research highlights the need for substantial changes in users' privacy practices and awareness to foster safer usage environments on social media sites.
When using Instagram, previous research [13] reported that more than half of the participants, or 62% out of 428 respondents, were actively sharing information.Moreover, these respondents felt that these platforms did not provide adequate privacy controls.The research also found that most active Instagram users did not fully understand the privacy policy and the terms of use applied on this platform.Besides that, there has been no research related to user behavioral intentions in using social media.Meanwhile, several studies have shown that behavioral intention factors are very important for social media users to share their private data [14].
Based on the limitations of previous research and the absence of research related to user behavioral intentions in using social media, this study aims to provide a breakdown of the variables that can affect social media confidentiality problems utilizing confirmatory factor analysis (CFA) and SEM to evaluate designs and verify theories, respectively.It is anticipated that the findings of this study will benefit the business and further studies.
The use of CFA and SEM in this research is strategically justified based on the advantages of each in achieving research objectives.CFA, a statistical technique, has proven invaluable in assessing the validity and reliability of measurement models by confirming the relationships between observed and latent variables.It is especially important in a complex area, such as users' behavioral intentions in using social media, to ensure a comprehensive understanding of the underlying constructs.SEM, on the other hand, offers a holistic approach by not only validating the measurement model through CFA but also simultaneously evaluating the structural relationships between variables.This integrated methodology is essential for differentiated analysis of the complex interactions between users' concerns, beliefs, awareness, and behavioral intentions.
Given the limitations identified in previous research and the specific focus on users' behavioral intentions in social media use, CFA and SEM appear as methodologically reasonable choices.These techniques offer a systematic and thorough exploration of variables, aligned with the overarching research goal of contributing valuable insights to the business world and paving the way for future research efforts.The comprehensive nature of CFA and SEM makes them indispensable tools for uncovering the complexities surrounding social media privacy and user behavior, making them a preferable choice over other methods that may be less comprehensive.

A. STRUCTURAL EQUATION MODELING
A statistical method, also known as SEM, uses multiple variables that combine various techniques, including CFA and regression (correlation).The objective is to investigate the connection among model parameters, constructions, or signals and their conceptual frameworks [15].Model solutions with at least one dependent factor can be solved using SEM.Since SEM relies on covariance examination, it produces a covariance matrix that is much more precise than that of linear regression analysis.The types of variables in SEM are divided into two: latent variables and observed variables.A latent variable is a variable that cannot be measured directly.This variable consists of several observed variables that will be tested first.Meanwhile, the observed variable is a variable whose variables can be measured directly [16].

B. CONFIRMATORY FACTOR ANALYSIS
The CFA analyses examination is a tool to determine how well variables that can be measured (indicators), which are also known as constructs in CFA, and describe or reflect an array of factors.A construct is an incalculable variable that needs to be described by variables that can be measured (indicators).The validation of the theory of measurement is also tested using CFA.How parameters are tracked is determined by the theory of measurement.It may clarify a construct shown in a model rationally and methodically [17].In this test, two stages are carried out: testing instrument validity and testing instrument reliability.
Instrument validity testing is carried out to determine the loading factor (LF) value of each indicator tested.The estimated LF value is not less than 0.5 and is considered ideal if it has a value of more than 0.7 [17].Instrument reliability testing is a test used to determine whether the indicators and variables used in research are reliable or consistent.This test is carried out by calculating the values of construct reliability and average variance extracted (AVE).A variable can be said to be reliable if it has a construct reliability value of ≥ 0.7.However, if the value on construct reliability is still in construct reliability ≥ 0.6 or construct reliability ≥ 0.7, then it is still allowed [16].Meanwhile, the calculation of AVE is optional; it can be adjusted to the needs of the research conducted [17].

III. RESEARCH FRAMEWORK
The research framework is proposed with the aim of understanding the extent to which users' awareness, privacy concerns, trust, and behavioral intentions can influence users' willingness to share personal data information through Instagram, TikTok, and YouTube social media.This research model was created by referring to the results of previous studies [2], [12] as shown in Figure 1.
Perceived privacy concern is defined as consumers' concerns about privacy and misuse of their personal data when using social media [18].Users who use privacy settings tend to maintain more social interactions, have more contacts on social networks, and provide higher-quality content [19].Privacy correlates with users' behavior in terms of sharing information publicly.Nowadays, most people are aware of the situation regarding privacy issues; nevertheless, they are not too concerned about doing so [2].
Perceived trust in this study refers to users' confidence when using social media.Trust is maintained if social media sites can act consistently with the interests and expectations that users want [20].When individuals have confidence in the functioning of social media sites, their desire to share personal information will increase [21].
User awareness can be defined as users' knowledge that leads to appropriate security behaviors.Knowledge alone is not enough.Awareness demands that the users behave according to their knowledge.Consistently poor security awareness is the root of major incidents, so it is important for individuals to be aware of the exchange of data on social media sites.
Behavior intention or individual desire is an intention or desire to utilize social media sites that are affected by a specific purpose.Meanwhile, individual desire is the extent to which an individual has constructed a feasible plan to do or not do something prescribed [22].Factors that influence behavioral intention in using social media are related to the goals that individuals want to obtain from social media use [23].
Based on the relationship of these variables, ten hypotheses are obtained in this model.
H1: Privacy concerns have a positive relationship with sharing information on social media sites.H2: User awareness has a positive relationship with sharing information on social media sites.H3: Perceived trust has a positive relationship with sharing information on social media sites.
H4: Privacy concerns are positively related to social media users' perceived trust.H5: User awareness is positively related to social media users' perceived trust.H6: User awareness is positively related to privacy concerns.H7: User privacy issues are positively related to users' wishes.H8: Users' desire is positively related to sharing data on social media sites.H9: The trust that users feel has a positive relationship with users' desire.H10: User awareness is positively related to the users' desire to use social media in the future.

IV. RESEARCH METHODOLOGY
The methodological steps in this research can be understood through Figure 2. The process began with identifying problems arising on social media sites.The next step involved a literature review, where several research articles related to social media and SEM were explored to gain in-depth understanding [24].After that, a survey questionnaire was created.This questionnaire was prepared with equal consideration to the number of variable indicators used.With reference to Figure 2, this process clearly visualizes the methodological flow of the research, from problem identification to the preparation of the survey questionnaire, ensuring clarity and consistency in each step of the research.
The demographics of respondents were Indonesians between the ages of 17 and 45 years who had experience using social media.These respondents were not restricted to gender, occupation, and income.A sample size of 391 respondents was decided as an ideal representation of a social media user population of around 215 million people.Random sampling and cluster sampling approaches are used to ensure that each individual in the population has an equal chance of being selected, and multiple groups or clusters reflecting variation in the population are drawn to increase the representativeness of the results.Although the focus is on social media users in Indonesia, this relatively small sampling is aimed at providing greater insight into the habits and perceptions of social media users at a global level, given the global nature of platforms such as Instagram, TikTok, and YouTube.
Data were collected by distributing forms online to potential respondents.Subsequently, data selection was conducted.Data that did not meet the requirements and were invalid were not used.Next, data were preprocessed utilizing two tests: validity and reliability tests.These tests ensure that the information tested in the study is true and trusted.The questionnaire is valid if the calculated R-value is ≥ the R table value with significance level of 5%, and it is reliable if Cronbach's alpha value is > 0.60 [24].Hence, it can be asserted that the respondents are consistent or stable from time to time.Both tests were carried out using the SPSS application.
Since the data used for further testing were normally distributed data, normality and a multicollinearity test were then conducted.The normality test is used to ensure that the data have a normal distribution with measurement criteria using skewness and kurtosis, as well as scatter plot diagrams.Skewness can be defined as the degree of skewness of a data distribution with a range of  3, while kurtosis is the degree of spiciness of a data distribution with a range of  8 [25].A .00 [25], [26].
CFA is a multivariate analysis used to test a concept built using several measurable indicators using the Amos application.There are two tests on the CFA test: the instrument validity test and the instrument reliability test.An instrument validity test is a test conducted by considering the loading factor (LF) value for every parameter used in the study.A good construct has a minimum LF value of 0.50.If the LF ≥ 0.50, it can be said that it is a valid indicator [26].In the instrument reliability test, the construct can be said to be good if the construct reliability score is ≥ 0.70 and the AVE value is ≥ 0.40 [27].After that, a modelfit test was performed to assess whether the variables being studied were normal by looking at several model-fit indices, chi-square (x 2 ), chi-square minimum divided by its degrees of freedom (CMIN/DF), goodness of fit index (GFI), adjusted goodness of fit index (AGFI), normal fit index (NFI), chisquare, comparative fit index (CFI), and root mean square error of approximation (RMSEA).If the model did not match, then the index modification was carried out on the model.This stage was carried out by adding the error covariance relationship based on the modification indicated in the SEM model output.
The next stage was to analyze the model with the tested hypothesis and to give suggestions and recommendations based on the results of the analysis obtained.

V. RESULTS
Table I shows the majority of the respondents were between 26 and 30 years old (34.3%).The remaining 24% were between 21 and 25 years old, 17% were between 31 and 35 years old, 13% were between 17 and 20 years old, 7% were between 35 and 40 years old, and 3.6% were between 41 and 45 years old.Furthermore, 63.9% of the respondents were female respondents, and the rest were male.Based on the education, bachelor's degree was the most dominant degree with 83.9%.Others included master's degree with 8.4% and high school with 7.7%.For jobs, the majority of the respondents worked as private employees (33.8%).The rest of the respondents were students (23.3%), entrepreneurs (13.3%), state-owned enterprise employees (8.4%), individuals in other occupations (7.7%), government officials (5%), individuals working in the health sector (4%), individuals working in the education sector (4%), and unemployed individuals (0.3%).Then, respondents' daily social media usage was more than 3 hours (65%).The rest used media social between 2 and 3 hours (22%) and less than 2 hours (12.8%).

A. MEASUREMENT MODEL
At the preprocessing stage, reliability and validity assessment were performed to ensure that the 19 questions of  the questionnaire were valid and reliable.The results showed that the calculated R score was bigger than or identical to 0.138, which is the critical R table value at a significance level of 5%.It indicates that the statistical analysis conducted in the study is statistically significant, as the calculated R-value surpasses the critical threshold, supporting the findings.Additionally, the followings are the validity test results for various constructs.First, privacy constructs used four different measures represented as PC1, PC2, PC3, and PC4.The numerical values assigned to these measures were PC1 (0.879), PC2 (0.850), PC3 (0.912), and PC4 (0.908).Importantly, all these values were bigger than the critical R table value for significance.It implies that the measures utilized to evaluate privacy in the study are valid and accurately represent the concept of privacy.Second is trust constructs.Trust is another construct examined in the study, assessed through five different measures represented as T1, T2, T3, T4, and T5.The numerical values assigned to these measures were T1 (0.873), T2 (0.861), T3 (0.812), T4 (0.861), and T5 (0.884).Same as privacy, all of these values surpassed the critical R table value for significance.It demonstrates the validity of the trust measures used in the study.Third is desire to share information on social media constructs.The study explored the desire to exchange data on social media through three measures denoted as BI1, BI2, and BI3.The numerical values for these measures were BI1 (0.874), BI2 (0.872), and BI3 (0.871).All these values were greater than the critical R table value, indicating the validity of these measures in assessing the desire to share information on social media.Fourth is awareness constructs.The awareness was assessed using five measures represented as AW1, AW2, AW3, AW4, and AW5.The numerical values assigned to these measures were AW1 (0.813), AW2 (0.837), AW3 (0.852), AW4 (0.774), and AW5 (0.782).Similar to the previous constructs, all of these values exceeded the critical R table value, affirming the validity of the awareness measures used in the study.Fifth is information sharing constructs.The study examined information sharing through two constructs represented as IS1 and IS2.The numerical values for these measures were IS1 (0.944) and IS2 (0.934).Both values significantly surpassed the critical R table value, indicating the validity of these measures in assessing information sharing.
The statistical analysis in the study was deemed significant, and all the constructs used, including privacy, trust, the desire to share information on social media, awareness, and information sharing, had been validated, as their numerical values exceeded the critical R table value for significance.It suggests that the study's findings are not only statistically meaningful but also that the measures employed accurately represent the concepts they are intended to assess.
Cronbach's alpha values was calculated for five crucial variables in a research study.Perceived privacy concern exhibited a high level of internal consistency with a Cronbach's alpha of 0.910, indicating the reliability of the questions measuring this construct.Similarly, perceived trust showed strong internal consistency with a value of 0.911, reinforcing the reliability of trust-related items.Behavioral intention demonstrated good internal consistency at 0.842, while user awareness scored 0.871 and information sharing scored 0.865, both indicating reasonably reliable measurements.
It can be said that the data is reliable if the Cronbach's alpha value is > 0.60 [24], meaning that the respondents of the questionnaire questions have been consistent or stable from time to time.All the examined concepts' Cronbach's alpha values were more than 0.60.It demonstrates the suitability of each concept employed in the experiment for the purpose of the study.Next was the normality test stage.This test can help determine whether the data are representative of all people or are regularly dispersed.The findings demonstrated that each model's spread of data was within its acceptable limits: skewness with a range of 3 and kurtosis with a range of 8.Hence, it can be said that the data are normally distributed.
Furthermore, in the multicollinearity test, the tolerance value was > 0.10 and the VIF value was < 10.00, so that there was no multicollinearity between variables.Then, it proceeded with validating the instrument by looking at the value of the LF.The LF value for each indicator was ≥ 0.5, so it can be concluded that the indicators for each variable were valid.The next stage was the instrument reliability test.
The description shows measurement to explain how well the aspects and indicators can represent the conditions being studied [27].In the CFA test, construct has good reliability if it has a construct reliability value of ≥ 0.70 and an AVE value of ≥ 0.40 [27].
The four indicators within the PC group, including PC1 to PC4, exhibited strong reliability in representing their respective aspects.PC1, characterized by a standard deviation (Std) of 0.833 and measurement error of 0.306, showcased a construct reliability of 0.911 and an AVE of 0.721, indicating its robustness.PC2 mirrors PC1's reliability metrics with an Std of 0.779 and measurement error of 0.393, maintaining a construct reliability of 0.911 and an AVE of 0.721.PC3 maintained this high level of reliability with an Std of 0.892 and measurement error of 0.204, alongside a construct reliability of 0.911 and an AVE of 0.721.Similarly, PC4 exhibited a construct reliability of 0.911 and an AVE of 0.721, with an Std of 0.887 and a measurement error of 0.213.Collectively, these indicators consistently demonstrate strong reliability in capturing the variance within their associated aspects.
The T indicators, including T1 to T5, consistently exhibited strong reliability in representing their respective aspects.T1, with an Std of 0.842 and measurement error of 0.291, boasted a construct reliability of 0.912 and an AVE of 0.674, surpassing established reliability thresholds.T2 followed suit, with an Std of 0.827 and measurement error of 0.316, maintaining a construct reliability of 0.912 and an AVE of 0.674, reaffirming its reliability.T3 maintained this high standard with an Std of 0.750, a measurement error of 0.438, construct reliability of 0.912, and AVE of 0.674, highlighting its effectiveness.Similarly, T4 showcased a construct reliability of 0.912 and an AVE of 0.674, with an Std of 0.827 and measurement error of 0.316, underlining its reliability.T5, with an Std of 0.854 and a measurement error of 0.271, aligned with its counterparts, maintaining a construct reliability of 0.912 and an AVE of 0.674, reinforcing its reliability in capturing the variance within its associated indicators.
The BI indicators, comprising BI1 to BI3, consistently met the criteria for reliability.BI1 maintained a construct reliability of 0.843 and an AVE of 0.642, supported by an Std of 0.802 and a measurement error of 0.357, underscoring its ability to effectively represent its aspect.Similarly, BI2 exhibited a construct reliability of 0.843 and an AVE of 0.642, confirming its reliability, complemented by an Std of 0.793 and a measurement error of 0.371, indicating its consistent capability in capturing the variance within its associated indicators.In parallel, BI3 achieved a construct reliability of 0.843 and an AVE of 0.642, emphasizing its reliability, with an Std of 0.809 and measurement error of 0.346, further highlighting its effectiveness in representing its aspect.
The AW indicators, encompassing AW1 to AW5, consistently demonstrated reliability in representing their respective aspects.AW1 showcased a construct reliability of 0.872 and an AVE of 0.577, supported by an Std of 0.748 and a measurement error of 0.440, emphasizing its capability.Similarly, AW2 maintained a construct reliability of 0.872 and an AVE of 0.577, confirming its reliability, complemented by an Std of 0.809 and measurement error of 0.346, consistently capturing the variance within its associated indicators.AW3 had a construct reliability of 0.872 and an AVE of 0.577, signifying reliability, with an Std of 0.819 and measurement error of 0.329, highlighting its effectiveness.Likewise, AW4 exhibited a construct reliability of 0.872 and an AVE of 0.577, confirming its reliability, supported by an Std of 0.704 and measurement error of 0.504, underscoring its ability.AW5 maintained a construct reliability of 0.872 and an AVE of 0.577, reaffirming its reliability, with an Std of 0.710 and measurement error of 0.496, consistently capturing the variance within its associated indicators.
The IS indicators, encompassing IS1 and IS2, consistently demonstrated strong reliability in representing their respective aspects.IS1 showcased a robust construct reliability of 0.867, accompanied by a high AVE of 0.765, underlining its effectiveness.Its Std of 0.882 and a measurement error of 0.222 further emphasize its capability in representing its aspect.Similarly, IS2 achieved a construct reliability of 0.867 and a high AVE of 0.765, confirming its reliability in capturing the variance within its associated indicators.With an Std of 0.867 and measurement error of 0.248, IS2 consistently maintained this high level of reliability.

B. STRUCTURAL MODEL
The next stage was a model suitability test, by looking at several model suitability indices (GFI), including chi-square, CMIN/DF, GFI, AGFI, CFI, Tucker-Lewis index (TLI) or nonnormed fit index (NNFI), and RMSEA, to comprehensively assess how aligned the SEM is with the data.The combination of these steps provides a different understanding of model adequacy by addressing different aspects of goodness of fit, such as absolute goodness of fit, relative goodness of fit, and parsimony.The results of this test are very important to determine whether the hypothesized model adequately represents the fundamental relationships between research variables.As a critical phase in the research process, model suitability testing assessed the adequacy of the proposed SEM in capturing the relationships between variables.This evaluation is important to ensure that the theoretical framework is aligned with the observed data.Several model fit indices, commonly referred to as GFI measures, were examined to measure the overall fit between the hypothesized model and the empirical data.
Table II presents the outcomes of the model-fit test, evaluating the congruence between the research model and the gathered data.Multiple model-fit indices were utilized to gauge the fit quality.The chi-square statistic yielded a value of 159.080, surpassing the recommended cutoff of 170.809, indicating a good fit between the model and data.The GFI was 0.959, exceeding the threshold of 0.9, signifying a favorable model fit.The AGFI stood at 0.945, within the acceptable range of 0.8 to 0.9, indicating a satisfactory fit.The CMIN/DF ratio was 1.120, below the cutoff of 2.00, demonstrating a good fit.
The CFI was 0.996, above 0.9 and close to 1.0, indicating a strong model fit.The TLI or NNFI also stood at 0.996, confirming a good fit.The RMSEA was 0.018, lower than the recommended threshold of 0.08, signifying a solid fit.In summary, across all indices and cutoff values, the research model was found to have a robust fit with the data, suggesting the model accurately represented the observed relationships.
After carrying out the model-fit test stage, the next stage was the analysis of the relationship between the variables and the hypothesis being tested.The hypothesis was analyzed by looking at the value of the critical ratio (CR) and the probability value (p-value) as proof of the truth of the hypothesis.The hypothesis was accepted if the CR > 1.96 and the p < 0.05.Table III shows the value of each variable relationship according to the hypothesis.
In the obtained structural path, the H1 hypothesis had a CR = 4.315 and p = 0.000, which backed up the theory used in this research.It demonstrates that an individual's motivation to disclose data through social media rises when they have influence over the privacy and security settings provided by social media sites because they feel safe on these sites and their confidence to share information increases.This finding is in line with the previous research [19].
The results of the analysis for the H2 hypothesis showed a value of CR = 2.719 and p = 0.007, supporting the H2 hypothesis.The less eager a person is to disclose data, the more conscious they are of the value of maintaining confidentiality and safety on social media platforms.People will typically use the web more cautiously in the future, particularly when using social media [17].
Furthermore, the results of the analysis for the H3 hypothesis showed a value of CR = 3.641 and p = 0.000, which also significantly supported the H3 hypothesis.This finding is in line with the previous research, when someone has a high trust in the social media sites used, the willingness to share personal information or data on social media will also increase [28].The results of the analysis for the H4 hypothesis showed a value of CR = 9.651 and p = 0.000, which significantly supports the H4 hypothesis.These findings indicate that if users have control over information and protect their social media privacy, it will increase their confidence in the ability of social media to protect their personal information.Similar findings were also reported in the previous study [2].
The results of the analysis for the H5 hypothesis showed a value of CR = 3.650 and p = 0.000, which supports the H5 hypothesis.This hypothesis confirms that an increase in awareness about privacy and security will also contribute to the increase in the level of users' trust in social media.Similar findings with previous research indicate that the higher the individual's awareness of social media privacy and security, the higher their level of trust in the site will be [29].It shows the importance of raising public awareness about privacy and security on social media.
Next is the results of the H6 hypothesis analysis, CR = 6.520 and p = 0.000 showed a significant positive correlation.It shows that users' awareness of privacy, level of trust, and their readiness to divulge sensitive data on social media increase because of their greater understanding.The higher the users' awareness of the importance of privacy, the more actively the users will take advantage of the privacy features provided by the sites they use, such as Instagram, TikTok, and YouTube.
The results of the analysis for the H7 hypothesis were CR = 3.855 and p = 0.000, which significantly support the H7 hypothesis.It shows that if users have control over access to their privacy data security, then their trust increases, and this causes them to use social media again.This finding is supported by the previous study [17].Since the site is considered capable of providing security for them, users will intend to use social media in the future.
The obtained results of the analysis for the H8 hypothesis were CR = 4.689 and p = 0.000, suggesting that it supports the H8 hypothesis.The users' intention or desire to reuse social media in the future has a positive effect on their desire to share information on social media.With the repetitive behavior, users know that the site used is safe, and this increases their desire to share information on the site used [30].
The obtained results of the analysis for the H9 hypothesis were CR = 5.609 and p = 0.000, which significantly support the H9 hypothesis.If the user believes in the security of the social media site used, the user's intention to reuse the site will increase.This is supported by the previous research [31].
The obtained results of the analysis for the H10 hypothesis were CR = 3.358 and p = 0.000, indicating a significant positive correlation.This finding is supported by the previous research, which has stated that users' intentions in using social media sites are influenced by perceived ease of use, social influence, perceived cost, trust, and awareness, so if users have a high awareness of the security of their information, users will also have wise behavior in using social media in the future [32].The result of a hypothesis test is shown in Figure 3.

VI. CONCLUSION AND LIMITATIONS
The findings of the study underscore the pivotal role of users' concerns, trust levels, privacy awareness, and behavioral intentions in shaping their willingness to share confidential data on social media platforms.Positive correlations emphasize the significance of users feeling secure about their private information, leading to increased engagement and active information sharing.Notably, heightened privacy awareness encourages users to use privacy features more frequently, fostering a cautious approach in their social media interactions.Empowering users with control over their data significantly enhances trust levels, contributing to an overall sense of wellbeing regarding private information and shaping users' intentions to sustain engagement with social media in the future.The study highlights a growing confidence among individuals in using social media and exchanging data, driven by increased awareness of confidentiality and safety.Users' confidence levels significantly influence platform loyalty and responsible data-sharing behavior.

TABLE III HYPOTHESIS TEST
The research's strengths lie in its meticulous methodology, utilizing SEM and CFA for a comprehensive exploration of complex dynamics governing users' behaviors on social media platforms.Emphasizing users' concerns, trust levels, privacy awareness, and behavioral intentions deepens the understanding of social media engagement.Findings underscore the critical impact of these factors on users' willingness to share data and the positive influence of heightened privacy awareness on responsible social media interactions.The study contributes by highlighting the role of user confidence in shaping platform loyalty and responsible data sharing.
However, the study's limited scope to Indonesia raises important considerations, excluding perspectives from users in other countries.The exclusivity to Indonesia may limit generalizability, raising questions about the universality of identified factors influencing users' social media behavior.To address this issue, future research is recommended to explore diverse aspects impacting individuals' motivations for sharing information on social media, encompassing social, economic, psychological, and other factors.Cross-national comparisons can yield a comprehensive understanding, informing the development of robust security measures.In addition, further investigation into nuanced principles guiding users from different backgrounds would offer a comprehensive understanding of social media engagement variations.Service providers are encouraged to enhance security offerings, ensuring a convenient and secure social media usage experience in the future.
Future study could involve investigating the longitudinal impact of heightened privacy awareness on users' social media behaviors.Tracking changes in user behaviors over time as privacy awareness initiatives are implemented would provide insights into the long-term effectiveness of such measures, contributing to the ongoing discourse on digital privacy and security.