Hackishness in DarkOwl Vision
Hackishness is defined as the rating of how likely content could be used for criminal activity.
How is Hackishness used in DarkOwl Vision?
A "Hackishness" rating is assigned to every piece of content collected by DarkOwl Vision, at the time of indexing. It is based on a machine learning algorithm that considers many different variables, such as patterns, metadata, terms, and more; no single element is responsible for the final score.
The Hackishness Algorithm
The algorithm defines a set of feature vectors with over 100 independent decision variables, see the table below for examples. The algorithm is probabilistic and takes into account both the presence and the absence of these variables.
The rating scale is from 0.01 (1%) to 1.0 (100%).
Examples of Feature Vectors |
Examples of Decision Variables |
|
Metadata |
Title, Domain, URL, Languages |
| Patterns | Social Security Numbers, Prose, Credentials, MD5 |
| Attribute Metrics | Number of Credit Cards or Email Addresses |
| Positive and Negative Terms | Hack, Dox, 0 day |
Case Studies Where Hackishness Has Been Used
- Username or Email + Password: Hackishness AND emailDomain
- People Selling Items (Credit Cards, Drugs, etc): Hackishness AND “credit cards”
- Database Dumps: Hackishness AND “data found”
- Large Volume Dumps (Credit Cards, Emails): Hackishness AND ccCount
- Pages Containing PII: Hackishness AND “dox”
- Government or Military Email Leaks: Hackishness AND emailDomain