This article is contributed. See the original author and article here.
Today, we are announcing the private preview for the classification improvements and would like you to join the preview and give us your valuable feedback.
Our team has been working on an exciting new update to the M365 S+C classification. The set of upcoming features would help you improve accuracy of your policies across M365 S+C solutions and increase coverage. We have improved accuracy of the out-of-the-box Sensitive information types, introduced confidence levels to simplify policy creation and added new capabilities to customize SITs which will allow you to fine-tune your SITs to reduce False positives & False negatives. In addition, we are adding new Sensitive information types that will allow you to increase your policy coverage.
Introducing Confidence levels
We are upgrading confidence levels by replacing number-based confidence system with three discrete confidence levels – high confidence, medium confidence & low confidence. Confidence levels (high, medium, low) reflect how much supporting evidence was detected along with the primary element. The more supporting evidence an item contains, the higher the confidence that a matched item contains the sensitive info you’re looking for. For example, matches with a high confidence level will contain more supporting evidence in proximity of the primary element and would lead to lower false positives but higher false negatives. Whereas matches with a low confidence level would contain little to no supporting evidence in proximity and thus would lead to zero to low false negatives but higher false positives.
We have made platform & definitional changes to help reduce False positives & False negatives. These improvements impact 80% of our Out-of-the-box sensitive information types. In early testing within Microsoft & with few of our design partners, we have seen a reduction in false positive rate from 30-60% to below 10% in some cases for the SIT pattern with the highest confidence level.
Copy & edit OOB SITs & other Custom SIT improvements!
We are addressing one of the top asks to allow copy & edit of out of the box sensitive type to fine-tune SITs based on your tenant needs to reduce false positives & false negatives. In addition, we have significantly improved the user experience & capabilities of creating a custom SIT. We have added two new power features to the custom SIT flow along with other improvements – Validators & additional checks. Validators will allow you to define your own regex and use validators like Luhn check. Additional checks will enable scenarios like excluding samples such as 4111111111111111 from credit card matches, define starting or ending characters, include/exclude prefix, etc. With the new set of advancements, you will be able to customize the SITs as per your requirements to further reduce False positives or false negatives.
100+ new & enhanced SITs
We have added 98 new SITs and enhanced definitions for 27 existing SITs. You can refer the list of new & enhanced SITs at the bottom of this email.
Sign up for preview program
If you are interested in start using any of the above features, you can join the preview program by filling the consent form & we’ll add you to our MIP&C Preview team on Teams.
Who can join?
These improvements will be rolling out to all M365 S+C solutions. However, in the initial preview phase, we will be enabling them for Communication compliance and DLP for SharePoint, OneDrive, Teams and Exchange. So, if you have active policies for any of these solutions, you can participate in the preview program.
The preview program will start in early-September and the features will light up in a phased manner. In the first phase you will have access to all the new & enhanced SITs, Improved accuracy and Confidence levels. Soon after, in the second phase, you will have access to Copy & edit OOB SITs & other Custom SIT improvements.
What is the expectation from you?
Create policies for sensitive information types you are interested in protecting and analyze the results to calculate the False positive & false negative rates. If you are seeing a false positive rate higher than 10% for the high confidence pattern or noticing false negatives, report it to us on the Class Depth Teams channel or write us an email at ClassDepthPreview@microsoft.com. We will help you bring them down by fixing issues on our side or by proposing customizations to your SITs.
Brought to you by Dr. Ware, Microsoft Office 365 Silver Partner, Charleston SC.