In this era of globalization, governmental institutions face significant challenges in maintaining their financial and operational integrity amidst increasingly complex economic and administrative dynamics. Uncertainty in public policies, demands for effective budget management, and the high need for accountability and transparency are primary focuses. Governments must ensure robust systems, both financially and in managerial practices, to prevent fraud amid financial instability. According to[1], Fraud Detection can be described as a series of actions aimed at identifying early signs or strong indicators of potential fraudulent activities, while limiting opportunities for perpetrators. Religiosity, as discussed by [2] and [25], relates to individuals' levels of conception and commitment to their religion. Individuals with high religiosity tend to have strict moral standards and strong work ethics, making them better able to identify fraudulent actions. CAATs, as defined by [4], are tools, technologies, and software aiding auditors in control testing, data analysis, and auditing. The use of CAATs can improve audit efficiency and effectiveness, allowing auditors to conduct tests and analyze data more accurately and swiftly. [5] suggest that auditors should no longer rely solely on conventional approaches in Fraud Detection, as failure to gather accurate audit evidence can significantly impact the occurrence of fraud. Task Specific Knowledge, according to [6], pertains to information relevant to audit tasks, aiding auditors in understanding the audited environment and internal conditions, thereby enhancing their assessment quality. [7] explain that Big Data refers to large, complex data sets requiring advanced technology for analysis. Govindan et al., (2018) state that Big Data has the potential to enhance forensic audit functions in Fraud Detection.

Previous studies by [9] dan [10] find that religiosity affects Fraud Detection by auditors, consistent with the research by [11] which asserts the significant influence of religiosity on auditors' ability to detect fraud. However, research by[12] suggests different results, indicating that religiosity does not influence internal auditors' ability to detect fraud. Findings from[13], [14], [15], dan [16] support the significant impact of CAATs on Fraud Detection. Research by [17] using a qualitative approach confirms that CAATs implementation plays a crucial role in the audit process and benefits Fraud Detection. Conversely, studies by [19] dan [18] suggest that information technology usage does not affect Fraud Detection by auditors. Studies by [20], Betri & Kusumawaty (2019), [22] indicate that Task Specific Knowledge influences Fraud Detection by auditors, supported by research by [23], [24], [25], dan [26]. [27], [28], [10], dan [29] have shown that Big Data influences Fraud Detection. However, research conducted by [30] yielded different results, indicating that Big Data does not affect Fraud Detection.

Religiosity plays a crucial role in enhancing auditor integrity during the audit process. Moreover, the utilization of Computer-Assisted Audit Techniques (CAATs) enables auditors to optimize the efficiency and effectiveness of audit procedures. This facilitates more accurate and swift data testing and analysis by auditors. Additionally, the pivotal role of Task Specific Knowledge cannot be overlooked, as it aids auditors in planning and executing audit procedures more systematically, thereby enhancing the quality of their assessments. Throughout the audit process, Task Specific Knowledge assists auditors in comprehending and resolving audit tasks, consequently improving the quality of their evaluations. Furthermore, Big Data exerts a significant influence, reinforcing the impact of auditor religiosity by providing broader access to relevant data. This enables auditors to track suspicious transactions that may contravene ethical or integrity principles. Moreover, in terms of CAATs utilization, Big Data enhances its effectiveness by enabling rapid and accurate analysis of large data volumes, thereby strengthening auditors' ability to identify suspicious patterns or anomalies. Lastly, Big Data also reinforces the influence of specific knowledge on audit tasks by providing broader access to relevant industry and business data. This allows auditors with in-depth knowledge of the audit context to identify suspicious patterns or trends that may go undetected manually.

[31] developed the Unified Theory of Acceptance and Use of Technology, which posits that technology usage is influenced by four main factors: performance expectancy, effort expectancy, social influence, and facilitating conditions. This theory provides insights into how information technology can aid in Fraud Detection by auditors by understanding the factors that influence technology adoption. The concept of cognitive dissonance theory, first proposed by [32], lays a crucial foundation for understanding communication dynamics and social influence. This theory highlights the inconsistency between cognitive elements, creating psychological discomfort. [26] underscore that cognitive dissonance theory impacts the change in auditor attitudes to predict intentions, aiming to reduce the arising inconsistency or dissonance. These cognitive factors guide auditors to ensure that the opinions they express after an audit align with the evidence they uncover during the examination process. The assumption of this theory is that individuals tend to seek consistency between beliefs, attitudes, and behaviors. Inconsistency among these components leads to cognitive dissonance, creating psychological discomfort. Thus, this theory provides insights into how Task Specific Knowledge can influence the Fraud Detection process conducted by auditors.

Attribution theory, first proposed by [33], addresses how individuals explain the causes of behavior and events. According to Fiske & [34], this theory discusses how social perception is used to make causal explanations about events. Its assumption is that individuals tend to explain behavior, whether their own or others', by attributing causes to internal or external factors, known as dispositional and situational attributions. Thus, this theory provides insights into how Auditor Religiosity can influence the Fraud Detection process. The way auditors interpret suspicious behavior will be influenced by religious values, which are dispositional attributions.

According to [35], fraud is defined as an unlawful act that can be committed by individuals, both internal and external to an organization, with the motive of personal or group gain, ultimately resulting in direct losses to others. In efforts to enhance performance and reputation, organizations may resort to illegal actions, disregarding the consequences and parties affected by their actions. Detection, as per the [36], refers to the effort to discover and determine facts, assumptions, or the existence of something. [37] defines Fraud Detection as a process of identifying the occurrence of fraud, its perpetrators, victims, and the reasons behind the incidents.

According to [2], religiosity is defined as the level of conception about religion and the level of commitment in religious practices. Similarly[3] explain that conception refers to an individual's understanding of their religious aspects, while commitment level relates to a comprehensive understanding in religious life. Hence, it can be concluded that individuals adopt different approaches to express and embody their religiosity. Individuals with high religiosity tend to exhibit strict moral standards and strong work ethics, thus being better able to identify fraudulent actions. [38] elucidate that religion has a binding nature, implying rules and responsibilities to be adhered to by its followers.[12] explains that high religiosity can enhance an auditor's independence as they tend to uphold truth and justice steadfastly.

According to [4], CAATs are the utilization of various tools, technologies, and software to assist auditors in performing various audit tasks, including control testing, confirmation, financial statement data analysis, as well as continuous auditing and monitoring. As per Auditing Standard Section 327 in computerized accounting information systems environments, manual testing is impractical for auditors. Therefore, computer-assisted audit techniques, or CAATs, are necessary to enhance auditor performance and efficiency and improve Fraud Detection effectiveness. [40] emphasize that CAATs enrich the audit process through technology utilization, employing specialized software that enables auditors to achieve audit objectives more efficiently.

According to [6], Task Specific Knowledge refers to information stored in memory, including experiences, practical facts, and theoretical concepts related to audit tasks, especially in evaluation and assessment processes. This information encompasses general and specific knowledge relevant to audit tasks. [25] states that Task Specific Knowledge assists auditors in understanding the audited environment and internal conditions, facilitating structured planning and execution of audit procedures. During audits, Task Specific Knowledge aids auditors in completing tasks more effectively, enhancing the quality of their assessments. [41] assert that the success of organizational audits in uncovering fraud, along with auditors' audit experience, is a significant element in potential Fraud Detection in each audit cycle and comprehensive risk assessment.

According to [42], Big Data refers to a vast collection of data with complex and varied structures. [8] assert that Big Data has significant potential in enhancing forensic audit functions for Fraud Detection. [10] cites one of the benefits of Big Data as its ability to assist companies in identifying fraud risks. Consequently, Big Data can support internal auditors in detecting fraud by expanding the coverage of essential data sources. This enables better analytical processes, ultimately enhancing the audit quality in identifying fraudulent activities.

This study employs an associative causal research design to investigate the impact of Auditor Religiosity, Computer Assisted Audit Techniques (CAATs), and Task Specific Knowledge on Fraud Detection, with Big Data acting as a moderating variable. Primary data were collected via questionnaires distributed to the Representative Offices of the State Development Audit Agency (BPKP) in Sumatera. The study's population consists of 876 auditors, with a sample size of 276 auditors determined using Slovin's formula at a 5% significance level, through probability sampling with simple random sampling methods.

The research adopts a Structural Equation Modeling (SEM) approach, utilizing Partial Least Squares (PLS) methodology via Smart PLS 4 software. This choice is made over Covariance-Based SEM (CB SEM) methods, such as AMOS or LISREL, due to several considerations. The dataset’s non-normal distribution aligns with PLS SEM’s robustness to deviations from normality, whereas CB SEM requires stricter assumptions. Additionally, the complex model, involving multiple variables and interaction effects, is better suited for PLS SEM, which can manage intricate models with latent variables and smaller sample sizes effectively. Although CB SEM could yield comparable results with larger datasets, PLS SEM’s flexibility and efficiency align better with the research's objectives and constraints.

In summary, employing Smart PLS 4 for SEM analysis ensures rigorous evaluation of both the outer and inner models. This includes assessments of convergent validity, discriminant validity, and internal consistency reliability, as well as analyses of R Square, effect sizes (F Square), Q Square, and significance testing (t-tests and MRA). This methodological choice supports the robustness and validity of the study’s findings, adhering to best practices in empirical auditing research.

The characteristics of the respondents in this study consist of gender, age, highest education level, functional auditor position (JFA), and length of employment. There were 120 male respondents (54.55%) and 100 female respondents (45.45%). The majority of respondents were aged between 21 and 40 years, with the highest number in the age range of 31-40 years (34.09%). In terms of highest education level, most respondents had a D4/S1 educational background (68.18%), followed by D3 (20.91%) and S2 (10.91%). The most common functional auditor position was Auditor Pertama (29.09%) and Auditor Muda (28.64%). The majority of respondents (61.82%) had been employed for more than 10 years.

Based on the table above, the average score of respondents regarding the indicators of the Auditor Religiosity construct reaches 26.03, which significantly exceeds the theoretical median value of 18. These results indicate that the majority of respondents have a significant level of religiosity, exceeding the median value used as a benchmark in the theoretical range. Similarly, the average score of respondents regarding the indicators of the Computer Assisted Audit Techniques construct reaches 70.32, which significantly exceeds the theoretical median value of 51. These results indicate that the majority of respondents have a much higher perception regarding the Computer Assisted Audit Techniques construct in the audit process compared to the theoretical median value. Furthermore, the average score of respondents regarding the indicators of the Task Specific Knowledge construct reaches 61.63, which significantly exceeds the theoretical median value of 42. This finding indicates that the majority of respondents have adequate knowledge to handle audit tasks. Additionally, the descriptive statistical test results indicate that the average score of respondents regarding the indicators of the Big Data construct reaches 68.92, which significantly exceeds the theoretical median value of 54. This indicates that the majority of respondents have a higher perception regarding the Big Data construct. Finally, the average score of respondents regarding the indicators of the Fraud Detection construct reaches 48.75, which significantly exceeds the theoretical median value of 33. This finding indicates that the majority of respondents have a high awareness and ability in Fraud Detection.

H1a: Auditor Religiosity has a significant influence on Fraud Detection

Based on the data analysis, it is confirmed that Auditor Religiosity has a significant influence on Fraud Detection. Attribution theory, proposed by [33], is used as the framework to explain the research findings. This theory considers how individuals explain behavior and events, either through dispositional attributions (related to the individual) or situational attributions (related to the context). In this context, auditors with higher religiosity tend to have more positive dispositional attributions towards ethical behavior and are more likely to perceive unethical behavior as wrong, thereby increasing their likelihood of detecting fraud. In synthesis, attribution theory suggests that auditor religiosity influences the dispositional and situational attributions used by auditors in understanding behaviors and events associated with accounting fraud. Higher auditor religiosity can enhance more positive dispositional attributions and more accurate situational attributions, thereby improving auditors' ability to detect accounting fraud. Religious auditors may be more sensitive to indications of fraud because they are more inclined to believe that ethical behavior should always be maintained, regardless of whether the situation is supportive or not. Therefore, religiosity not only affects auditors' intrinsic motivation but also their assessment and reaction to potentially fraudulent situations. These findings are consistent with the research conducted by [9] and [10], which demonstrated the influence of religiosity on fraud detection by auditors. This research aligns with the study by [11], which showed that religiosity significantly impacts auditors' ability to detect fraud.

H1b: CAATs has a significant influence on Fraud Detection

Based on the data analysis, it is confirmed that CAATs (Computer Assisted Audit Techniques) do not have a significant influence on Fraud Detection. This is due to the fact that, although auditors have a positive perception of using Excel as an audit tool, the capabilities and complexity of Excel may not be sufficient to detect more sophisticated fraud. The Unified Theory of Acceptance and Use of Technology (UTAUT) is used as the framework to explain the relationship between the use of CAATs and fraud detection. UTAUT considers four main constructs: performance expectancy, effort expectancy, social influence, and facilitating conditions. In this context, although auditors have positive perceptions (high performance and effort expectancy), feel supported by colleagues (social influence), and have access to the necessary resources (facilitating conditions), the use of Excel remains inadequate for detecting more sophisticated fraud. This is due to Excel's limitations in analyzing highly complex data or detecting subtler fraud patterns, which often require more specialized and advanced audit tools. Furthermore, UTAUT helps explain that, although auditors perceive benefits and ease of use with Excel, and there is social support and adequate conditions, these factors do not automatically translate into increased effectiveness in fraud detection. This indicates that, beyond technology adoption, the quality and suitability of the audit tool for specific tasks such as fraud detection are also crucial. These findings are consistent with the research conducted by [19] and [18]. However, studies by [13], [14], [15], dan [16], which show that CAATs do not have a significant impact on Fraud Detection, present different results. The study by Widuri & [17], using a qualitative approach, also contrasts with this, indicating that the implementation of CAATs plays a crucial role in Fraud Detection.

H1c: Task Specific Knowledge has a significant influence on Fraud Detection

Based on the data analysis, it is confirmed that Task Specific Knowledge has a significant influence on Fraud Detection. This is attributed to two main factors: first, auditors with specialized knowledge in audit tasks tend to be more competent in identifying and analyzing signs of fraud. Second, in-depth knowledge of audit procedures and fraud detection techniques allows auditors to more accurately evaluate and interpret suspicious data. In the context of cognitive dissonance theory proposed by [32], to reduce this dissonance, individuals tend to seek consistency through changes in their attitudes, beliefs, or behaviors. Auditors with specific knowledge of audit tasks have a strong knowledge base to detect fraud. When faced with information or data that is inconsistent with their understanding of how financial processes should occur, they experience cognitive dissonance. To reduce this dissonance, auditors are more likely to delve deeper and conduct more thorough examinations to identify the source of these inconsistencies, ultimately enhancing their ability to detect fraud. These findings support previous research by [20], Betri & Kusumawaty (2019), [22], which showed that Task Specific Knowledge influences Fraud Detection by auditors. This research is supported by studies by [23], [24], [3], dan [26], and is consistent with cognitive dissonance theory.

H2a: Big Data Moderates the Influence of Auditor Religiosity on Fraud Detection

Based on the analysis results, it is confirmed that Big Data does not moderate the influence of Auditor Religiosity on Fraud Detection. This is attributed to two main factors: first, although Big Data offers significant potential to enhance data analysis capabilities, its effective use in fraud detection requires specialized tools, skills, and deep understanding that are not possessed by all auditors. Second, religiosity factors are more intrinsic and related to the personal values of auditors, which may not be directly influenced by sophisticated data analysis tools like Big Data. Attribution theory by [33] and UTAUT by [31] can help explain why Big Data does not moderate the influence of auditor religiosity on fraud detection. Religiosity is a dispositional factor that influences the personal values and ethics of auditors, while Big Data is a technical tool. Although Big Data can enhance data analysis capabilities, the religious values of auditors tend to influence how they perceive and react to fraud situations more than just the tools they use. Therefore, while Big Data can assist in fraud detection, the influence of auditor religiosity on fraud detection is more intrinsic and not significantly moderated by the use of Big Data.

H2b: Auditor Religiosity Moderates the Influence of CAATs on Fraud Detection

Based on the analysis results, it is confirmed that Big Data moderates and strengthens the influence of Computer Assisted Audit Techniques (CAATs) on Fraud Detection. This is attributed to two main factors: first, Big Data provides volume, variety, velocity, veracity, and value of data, enabling CAATs to analyze data on a much larger and more complex scale. Second, the integration of Big Data with CAATs allows auditors to detect patterns and anomalies that may not be visible with traditional audit techniques or smaller datasets. UTAUT by Venkatesh et al., (2003) can explain the relationship between these variables. With the existence of Big Data, auditors can conduct data analysis on a larger scale and at a faster pace. This enables CAATs to identify more complex and hidden fraud patterns that may not be detected with smaller datasets. The integration of Big Data and CAATs also enables auditors to perform predictive analysis and identify potential risks before fraud occurs. This enhances auditors' ability to proactively detect and prevent fraud. Thus, Big Data not only enhances the effectiveness of CAATs but also transforms auditors' approach from reactive to proactive in fraud detection.

H2c: Auditor Religiosity Moderates the Influence of Task Specific Knowledge on Fraud Detection

Based on the analysis results, it is confirmed that Big Data does not moderate the influence of Task Specific Knowledge on Fraud Detection. This is attributed to two main factors: first, although Big Data offers significant potential in data analysis, its effectiveness in detecting fraud largely depends on auditors' ability to utilize the data effectively. Auditors with specific knowledge about audit tasks may not yet be fully skilled or experienced in using Big Data, thus the potential of this technology is not fully realized. Second, Task Specific Knowledge tends to focus more on technical skills and knowledge directly related to the audit process, which may not be fully enhanced solely by the presence of Big Data without adequate data analysis skills support. By utilizing UTAUT by [31] and cognitive dissonance theory by Festinger (1957) as the conceptual framework, we can understand that although auditors possess strong Task Specific Knowledge, the success of Big Data utilization in fraud detection largely depends on how comfortable and competent they feel in using the technology. Performance and effort expectations, social influence, as well as facilitating conditions all play crucial roles in determining the extent to which Big Data can influence fraud detection. Discomfort or lack of skills in using Big Data may hinder the potential of this technology, even though auditors have adequate technical knowledge.

Based on the data analysis, several key conclusions and practical recommendations regarding fraud detection in auditing emerge. The study confirms that Auditor Religiosity significantly influences Fraud Detection. Auditors with higher levels of religiosity are more likely to adhere to strong ethical standards, which enhances their ability to detect fraudulent activities. This finding emphasizes the importance of integrating auditors' personal values and beliefs into their professional conduct. Organizations should thus foster an environment that supports and respects auditors' ethical standards and personal values.

Conversely, Computer Assisted Audit Techniques (CAATs) do not have a significant influence on Fraud Detection in this study. Although CAATs are intended to improve audit efficiency and accuracy, their effectiveness in detecting fraud may be limited by factors such as data quality and auditors' proficiency with these tools. It is essential for organizations to ensure that auditors receive comprehensive training in CAATs and that these techniques are properly implemented to maximize their potential benefits.

Task-Specific Knowledge, however, does significantly impact Fraud Detection. Auditors with specialized knowledge in their audit tasks are better equipped to identify and analyze fraudulent activities. This underscores the necessity for ongoing professional development and targeted training programs to keep auditors updated on the latest fraud detection techniques and best practices.

The role of Big Data in this context is nuanced. The study finds that Big Data does not moderate the relationship between Auditor Religiosity and Fraud Detection or between Task-Specific Knowledge and Fraud Detection. However, Big Data does moderate the influence of CAATs on Fraud Detection. This suggests that while Big Data itself may not directly enhance fraud detection, its interaction with CAATs can be beneficial. Organizations should therefore focus on improving their capabilities to manage and analyze Big Data effectively, leveraging it to enhance the functionality of CAATs.

In conclusion, to enhance fraud detection capabilities, organizations should invest in comprehensive training and development programs to improve auditors' task-specific knowledge and ethical awareness. Additionally, adopting and effectively integrating CAATs into the audit process, along with creating an ethical work environment that aligns with auditors' personal values, can significantly boost fraud detection efforts. Addressing these recommendations will help organizations overcome challenges associated with Big Data and optimize the use of auditors' attributes and technological tools in their fraud detection practices.