Abstract

The growing dependability of financial institutions on cloud-based services for document management has given rise to significant safety concerns over the potential for malicious PDF documents. In response, this research suggests a Machine Learning (ML) based detection technique for spotting erroneous PDF files in a financial cloud environment. The recommended solution leverages modern technologies to enhance the security posture of financial institutions by efficiently recognizing and mitigating potential hazards. The methodology analyzes the attributes of malicious PDF documents and the final relevant qualities that indicate malicious intent using the Sparse K-Nearest Neighbor (SKNN) algorithm. The PDF files are pre-processed to ascertain whether the samples have errors or duplicate content. Principal Component Analysis (PCA) is used to extract the features of PDF files. A user employs files that can avoid detection by a security system and send a variety of payloads that have the potential to do serious damage. Comprehensive testing and assessment in terms of ROC-AUC (98.7%), F1-score (98%), recall (97%) and detection accuracy (98.5%) demonstrate the effectiveness of the recommended strategy. The finding's outstanding performance demonstrates how ML techniques can be used to improve cloud financial system cybersecurity safeguards.