Researchers Map Method for Spotting Suspicious Insiders
Mining of email data could help companies spot dangerous employees before they do damage
![](https://eu-images.contentstack.com/v3/assets/blt6d90778a997de1cd/bltc94608acf452fd67/655cf371ab171e040a838b2a/329050_DR23_Graphics_Website_V5_Default_Image_v1.png?width=1280&auto=webp&quality=95&format=jpg&disable=upscale)
A group of researchers has discovered a way to identify potentially dangerous employees by analyzing the content of their email.
Three researchers at the Air Force Institute of Technology -- James Okolica, Gilbert Peterson, and Robert Mills -- have published a paper that outlines an algorithm for mining email data and identifying patterns of transmission that might tell managers when employees are keeping a secret.
In a nutshell, the algorithm identifies email topics of interest that are communicated outside the organization, but never shared with others inside the organization. The identification of such topics indicates that employees "either have a secret interest in the topic or generally feel alienated from the organization," the paper says.
In the study, researchers applied a data mining concept called Probabilistic Latent Semantic Indexing (PLSI), which has been used to extract specific information from a large body of data. By adding users to the body of data being studied, the researchers were able to identify patterns of content exchanged between specific users.
The researchers then tested their concept on the body of email messages left by the now-defunct Enron Corp. They identified a number of employees who might have been engaged in dangerous insider behavior prior to the company's fall.
The researchers used PLSI to extract specific topics from the email data, and then diagrammed "social networks" revolving around those topics. In most cases, the social networks occurred inside the organization. But in a few cases, the social networks revolved around a single individual inside the organization, who discussed the topic only with individuals on the outside. Such a network suggests that this individual might be keeping secrets from co-workers, they said.
The algorithm, dubbed the Potential Insider Threat Detection Algorithm, is a "promising tool" for aiding IT departments in narrowing down the list of subjects in a breach investigation, the researchers said. However, the experimental analysis of Enron's email did not correctly identify the top managers who were involved in the company's fraud.
"This may be because any revealing emails would have been to other people inside the organization," the paper says. "To overcome this, work needs to be done to extract insider threat collusion networks that involve a small number of individuals."
The researchers also hope to expand their study to include Internet activity, which would help identify Web browsing trends that might suggest topics of interest or behavior threatening the organization's security.
Have a comment on this story? Please click "Discuss" below. If you'd like to contact Dark Reading's editors directly, send us a message.
About the Author
You May Also Like
State of AI in Cybersecurity: Beyond the Hype
October 30, 2024[Virtual Event] The Essential Guide to Cloud Management
October 17, 2024Black Hat Europe - December 9-12 - Learn More
December 10, 2024SecTor - Canada's IT Security Conference Oct 22-24 - Learn More
October 22, 2024