Sentiment Analysis to Counter the Insider Threat

What Is the Insider Threat?

Nowadays organizations have to be alert to the problem of the insider threat. The most famous recent example is probably that of the contractor, Edward Snowden, who gained access to and stole highly classified documents while working as a highly cleared IT administrator at a National Security Agency facility.

The notion of the insider threat, however, is not just a problem for an intelligence organization. It applies to many different types of organizations with highly sensitive data:

• Financial institutions that must guard against insider embezzlement or other kinds of fraud
• Civilian Government agencies that need to protect citizens’ PII
• High-tech firms that want to protect their intellectual property against a rogue employee who might steal and sell it to a competitor.

Organizations need to take steps to guard against an employee or contractor, or anyone with access privileges to an organization, engaging in nefarious activity.

Sentiment Analysis Discovers Clues to Employees’ Attitudes in Unstructured Text Data

Happily, there’s a technology, Sentiment Analysis, which can be effectively used to help counter these threats.

Sentiment Analysis is used to uncover the attitudes towards a wide variety of topics expressed by employees in a wide range of forums, both internal and external to the enterprise:

• Email
• Social media (Twitter, Facebook, etc.)
• Blogs
• Web forums
• etc.

Sentiment Analysis identifies opinions and intentions by using techniques derived from AI and Machine Learning. It performs text analytics of the contents of these sources, which are all unstructured text data, looking for specific sentiments.

Sentiment Analysis is also able to process at scale, easily handling very large amounts of data, thus exceeding by many orders of magnitude the capabilities of purely human-driven monitoring.

How Sentiment Analysis Works

There are different types of Sentiment Analysis tools on the market today that define sentiment in various ways and use different approaches:

• In the simplest approach, positive phrases like “great” or negative ones like “terrible” are used to indicate the entire sentence or text is to be classified as positive or negative.
• Some Sentiment Analysis tools are also capable of introducing a multiplier factor: if a word like “very” or “really” occurs before the “great” or “terrible,” they use this information to quantify the strength of the sentiment.
• A critical later advance is the ability to connect a sentiment expressed in text to a specific item in the context, i.e., feelings concerning particular products, persons, or really anything at all. This more advanced technology is called Entity-Based Sentiment Analysis.

In this way Sentiment Analysis has rapidly become a highly important tool across a range of industries, helping, for example, product companies to understand customer responses to their products or marketing companies to gauge how well a marketing campaign is going. It is also highly effective in identifying potential insider threats among an organization’s employee population.

How Sentiment Analysis Can Be Used to Counter the Insider Threat

Organizations that perform vital tasks such as a financial institution or an intelligence organization can use Sentiment Analysis to look for potential insider threats. An employee might be expressing strong negative opinions or even anger about any aspect of an organization on a variety of platforms. These could touch on anything – from working conditions to benefits and even to doubts about an organization’s mission.

Here are examples of how Entity-Based Sentiment Analysis produces a structured representation of the meaning of the sentences below:

• “Health benefits are way inferior to my last place!”
  • sentiment: negative
  • entity: health benefits
• “I do not agree with the mission of this agency”:
  • sentiment: negative
  • entity: mission

In each example, the sentiment expressed is assessed as NEGATIVE. Each sentiment is also linked to objects of sentiments (Entities) that employees express opinions about (examples here are “health benefits” and “mission”). Obtaining information on WHAT an employee might be unhappy about is critical in identifying possible insider threats among other negative sentiments. Sentiments about an organization’s mission are obviously more relevant to the insider threat problem than those about health benefits.

Sentiment Analysis allows identification of individual employees who might be on a bad path.  If an employee expresses negative sentiments in an inappropriate way or at a rate far exceeding the average, it might be indicative of a potential insider threat.

Sentiment Analysis also needs to be used to identify actual direct threats uttered by employees, whether against other employees or against just about anything – public figures, organizations, etc. Any of those would need close human review by experts to assess their significance.

Summary

In sum, Sentiment Analysis is a valuable tool for protecting against the insider threat. Based on advanced AI techniques, it is an effective means of screening for those employees who might present a threat to an organization.