Recently, the adoption of Robotic Process Automation (RPA) solutions for business automation has significantly improved corporate productivity, but it has also increased the risk of security threats and information leakage caused by insiders. Existing...
Recently, the adoption of Robotic Process Automation (RPA) solutions for business automation has significantly improved corporate productivity, but it has also increased the risk of security threats and information leakage caused by insiders. Existing digital forensic research has primarily focused on file systems, network traffic, and memory dump analysis, while methodologies for tracking and analyzing automation activities performed using RPA tools have received little attention.
This study proposes a forensic approach that treats workflow definition files and execution logs generated by RPA tools as primary digital evidence to interpret the intent and execution traces of RPA automation activities. A dedicated tool was also implemented to automate the analysis process. The commercial RPA solution UiPath Community Edition and the open-source frameworks Robot Framework and taskt were selected as analysis targets. Their file storage structures, log generation mechanisms, and security vulnerabilities were comparatively analyzed. UiPath generates process definition files in .xaml format with text-based execution logs, Robot Framework produces .robot script files and output.xml process definition files, and taskt generates .xml process definition files along with text-based execution logs. Each tool exhibited distinct analytical characteristics in its file structures and logging systems. In the experimental phase, various scenarios representative of real-world environments were designed, including sensitive information exposure, automated email transmission, FTP file uploads and downloads, and timestamp manipulation. Analysis of the generated workflow definition files and execution logs enabled the identification of key digital evidence, such as sender and recipient information, attachment file paths, FTP server addresses, and authentication credentials. The results demonstrate that workflow definition files are valuable forensic artifacts for tracing automation activities, reconstructing data, and inferring the intent of RPA automation designers.
Based on these findings, a Python-based tool named “RPA WorkFlow Analyzer” was developed. The tool automatically identifies the RPA platform by examining the internal structure of input files and applies tool-specific parsing logic to extract activity lists, keywords, command structures, and attribute values. Through the analysis of collected workflow definition files and execution logs, automation activities could be retrospectively reconstructed, empirically validating the feasibility of digital evidence extraction and analysis in RPA environments.
This study pioneers the emerging field of RPA forensics and systematically demonstrates the forensic value of workflow definition files and execution logs. Although the experiments were conducted in a controlled environment using three RPA tools, which limits generalizability, the study provides a foundational methodology for RPA forensics. To address the limitations of static analysis in handling obfuscation and encryption, future research should incorporate memory forensics and dynamic analysis techniques. Furthermore, extending the research to cloud-based RPA platforms and investigating anti-forensic countermeasures, such as encrypted credential decryption, remain important directions for future work.