With the rapid evolution of malware and the limitations of traditional anti-virus techniques, machine learning has emerged as a promising tool for detecting new malware variants. This analysis explores the use of Elastic's outlier detection functionality, leveraging byte histogram profiles from binaries to identify potentially malicious software. The EMBER dataset, comprising features from 1.1 million Portable Executable files, serves as the foundation for experiments in classification and outlier detection, employing an ensemble of established outlier detection algorithms. The results indicate that byte histograms can highlight irregularities in malicious binaries compared to benign ones, particularly in the presence of obfuscation techniques that obscure ASCII characters. Although the study shows promise, it also underscores the need for further refinement in feature selection and threshold determination to enhance detection accuracy, demonstrating the potential of machine learning as a valuable asset for malware analysts.