Artificial Intelligence (AI) is used across the length and breadth of Securly products to varying degrees. Detection of suspicious activity on various platforms such as social networking sites, search engines, emails, or GDocs needs to be handled differently as suits the platform’s requirements. This can make understanding of how Securly uses AI a little confusing.
The matrix below simplifies this. Here are a few things to remember when reading the matrix:
- AI = AI used for that specific category for that product or feature
- N = category not supported for the corresponding product
- Phrase Match = searched phrase is matched to phrases in our database
- Keyword Match = searched keyword is matched to keywords in our database
- Domain Match = searched/browsed domain is matched to the domains in our database
- PoC = Successful at a PoC level, under development
This means that AI is used to detect the following*:
- Grief, Bully, and Violence events by the Auditors in Gmail.
- Grief, Bully, and Violence events by the Auditors in Google Docs.
- Grief, Bully, and Violence events on Social Networking sites such as Facebook and Twitter.
- Nudity Detection via Filter for URL Filtering (PageScan).
The method of keyword matching is used in the following scenarios*:
- Detecting the use of Profanity in Gmail and GDocs (Auditor).
- Detecting Pornography, Gambling, and Drugs in Search (Filter). Keywords indicating profanity are also supported but include overlaps with other categories such as bullying or nudity.
- Detecting Profanity on Social Networking sites such as Facebook and Twitter.
The method of domain matching is used to detect and block websites for almost all of these categories. A total of 14 categories are available for the IT admin to block per policy. The admin can choose to receive alerts for any of those 14 categories with one simple step via the Securly portal in the Policy Editor. You can also assign monitoring of these alerts via the Delegated Reporting Groups.
Nucleus UI view
Older UI view
Radicalization is handled separately where Securly partners with the Counter-Terrorism Internet Referral Unit (CTIRU) to detect suspicious activity using the domain matching method.
This method of phrase matching is used to detect Grief in searches for Filter users.
The rest of the categories are not covered by AI at this time, but research and PoCs are underway for almost all of them. We will keep you updated as things take shape.
*at the time of this writing.