SC-401 · Domain 1: Implement information protection · 34% of the exam
Task 1.1: Implement and manage data classification
Turning sensitive information requirements into built-in or custom sensitive info types, document fingerprinting, exact data match, trainable classifiers, OCR, and monitoring classification and label use in the explorers.
Study it
Sensitive info types: built-in, custom, fingerprinting and exact data match
Trainable classifiers, OCR, and the Content and Activity explorers
Lesson coming
Sample questions
Try each one before opening the answer. Every option is explained, with the Microsoft documentation page that proves it.
Question 1 · choose 1
Legal uses one standard nondisclosure agreement form. It holds the blank Word template and a dozen signed copies, and wants a DLP policy to detect filled-in copies of the form in outgoing email and in SharePoint. No other contracts may be matched. Which classifier should you build?
AA document fingerprint SIT created from the blank NDA template
BA custom trainable classifier seeded with the dozen signed copies
CAn exact data match SIT built from a table of the signers' names
DA custom SIT with a regular expression for the form's title
Show the answer and why
AA document fingerprint SIT created from the blank NDA template
Correct
A fingerprint is built from a standard form and detects documents created from that template, including filled-in copies.
BA custom trainable classifier seeded with the dozen signed copies
Incorrect
A custom trainable classifier needs at least 50 positive and 150 negative samples; a dozen copies can't seed it.
CAn exact data match SIT built from a table of the signers' names
Incorrect
EDM looks for values from a source table, not for documents that follow a form's wording.
DA custom SIT with a regular expression for the form's title
Incorrect
A title pattern matches any document that mentions the title, not only copies of the form.
Document fingerprinting converts a standard form into a SIT that DLP rules can use in Exchange, SharePoint, OneDrive, Teams and devices; it matches documents created from that template. A custom trainable classifier is seeded with at least 50 positive and 150 negative samples chosen by a person, so a dozen signed copies aren't enough to create one.
Contractor IDs are eight digits: the first two are a site code and the last six are the contract start date in YYMMDD format. A custom SIT with the pattern \d{8} matches thousands of random numbers. You must keep only numbers whose last six digits form a valid date. What should you add to the pattern?
AA checksum validator with weights for all eight digits and a mod value
BA Luhn check validator
CA date validator with the IDs' date format and a start offset of 2
DAn Exclude duplicate characters additional check
Show the answer and why
AA checksum validator with weights for all eight digits and a mod value
Incorrect
A checksum validator compares a calculated check digit with one position; these IDs have no check digit.
BA Luhn check validator
Incorrect
The Luhn validator tests numbers built with the Luhn algorithm, which these IDs don't use.
CA date validator with the IDs' date format and a start offset of 2
Correct
The date validator tests whether an embedded date is valid; the offset marks where the date starts in the match.
DAn Exclude duplicate characters additional check
Incorrect
This check drops matches whose digits are all the same; random numbers with valid-looking digits still match.
When a date is embedded in a number that a SIT pattern matches, a date validator checks that part: you choose the date format and the start offset, which is 2 here because the date follows the two-digit site code. Checksum and Luhn validators test check-digit algorithms, and Exclude duplicate characters removes only matches made of one repeated digit.
An admin edits a custom SIT's regular expression in the rule package XML that she uploaded in March, keeps <Version major="1" minor="0" build="0" revision="0" />, saves the file as Unicode and runs Set-DlpSensitiveInformationTypeRulePackage in Security & Compliance PowerShell. The cmdlet returns no error, but two days later the SIT still matches as before. What's the cause?
AThe rule package must be saved with UTF-8 encoding, not as Unicode
BThe Version element wasn't increased, so the new rule package wasn't deployed
CRule packages must be imported with Exchange Online PowerShell
DSet-DlpSensitiveInformationTypeRulePackage only adds new SITs
Show the answer and why
AThe rule package must be saved with UTF-8 encoding, not as Unicode
Incorrect
The rule package is saved as an .xml file with Unicode encoding, as she did.
BThe Version element wasn't increased, so the new rule package wasn't deployed
Correct
Microsoft 365 notes the version of the first upload; an updated package must carry a new version number or it isn't deployed.
CRule packages must be imported with Exchange Online PowerShell
Incorrect
Custom SIT rule packages are imported with Security & Compliance PowerShell; Exchange Online PowerShell no longer imports them.
DSet-DlpSensitiveInformationTypeRulePackage only adds new SITs
Incorrect
This cmdlet imports updated XML into the existing rule package.
When you first upload a rule package, Microsoft 365 records its Version element. If you update the package and upload it again without increasing the version number, the package isn't deployed. Updated XML is imported into the existing package with Set-DlpSensitiveInformationTypeRulePackage, in Security & Compliance PowerShell.
You're creating an exact data match SIT in the new experience from a CRM table with the columns AccountNumber, CustomerName, City and Phone. Account numbers are always CA- followed by eight digits, are unique per row, and no built-in SIT detects them. What should you do for the primary element?
AUse CustomerName as the primary element and map it to the All Full Names SIT
BMap AccountNumber to a custom SIT that matches any number of six to twelve digits
CMark AccountNumber as searchable and leave it without a SIT mapping
DCreate a custom SIT for CA- followed by eight digits and map AccountNumber to it
Show the answer and why
AUse CustomerName as the primary element and map it to the All Full Names SIT
Incorrect
Primary elements should make each row unique; names repeat across rows, which slows lookups and can cause time-outs.
BMap AccountNumber to a custom SIT that matches any number of six to twelve digits
Incorrect
A loosely defined SIT triggers lookups on many unrelated strings, which leads to false negatives and poorer performance.
CMark AccountNumber as searchable and leave it without a SIT mapping
Incorrect
EDM searches content for the primary element through a SIT; each primary element must be mapped to one.
DCreate a custom SIT for CA- followed by eight digits and map AccountNumber to it
Correct
Primary elements must be discoverable through a SIT, and a strongly defined pattern keeps lookups narrow.
EDM searches content for the primary element through the SIT it's mapped to, so primary elements must be discoverable by a SIT and follow a defined pattern. The new experience warns about primary fields mapped to loosely defined SITs, which cause lookups on unrelated strings, and about primary values repeated across many rows. A strongly defined custom SIT on a unique column avoids both.
Your custom trainable classifier Supplier quotes was published two months ago and now flags many internal price lists. Its creator wants it to stop matching price lists while still finding quotes. What should she do?
AMark the price lists as Not a match in Data explorer, so that the published classifier retrains on that feedback
BAdd the price lists to the classifier's negative seed folder and run a retrain
CCreate a new classifier with larger seed sets, price lists among the negatives, in dedicated SharePoint folders
DCreate a new classifier whose negative samples are the price lists in a folder of her OneDrive
Show the answer and why
AMark the price lists as Not a match in Data explorer, so that the published classifier retrains on that feedback
Incorrect
Retraining custom classifiers is no longer supported; feedback doesn't retrain a published classifier.
BAdd the price lists to the classifier's negative seed folder and run a retrain
Incorrect
Retraining custom classifiers is no longer supported; a new classifier is needed.
CCreate a new classifier with larger seed sets, price lists among the negatives, in dedicated SharePoint folders
Correct
To improve a custom classifier you remove it and start fresh with larger sample sets; seed sets go in dedicated SharePoint folders.
DCreate a new classifier whose negative samples are the price lists in a folder of her OneDrive
Incorrect
Seed content must be in SharePoint folders, not in a OneDrive folder or a Teams folder.
Retraining custom trainable classifiers is no longer supported: to improve one, you remove it and begin fresh with larger sample sets. Positive and negative seed content goes in separate SharePoint folders dedicated to the seed content, on a communication site or another SharePoint site type, not in a Teams folder or a OneDrive folder.
You're creating an exact data match SIT in the new experience. The source table has MemberID (the primary element, written as 1234-5678-90 in the table but often without dashes in documents), HolderName (values such as Ana Bowman) and BirthDate. Which two settings should you choose? (Choose TWO.)
AIgnore the dash as a delimiter for the MemberID column
BNarrow mode, so supporting elements from all primary matches in range count
CUpload the table unhashed so EDM can compare spelling variants
DMulti-token matching for HolderName
ESplit HolderName into two columns, because EDM can't match values that contain spaces
Show the answer and why
AIgnore the dash as a delimiter for the MemberID column
Correct
Configurable matching can ignore delimiters and punctuation in searched values.
BNarrow mode, so supporting elements from all primary matches in range count
Incorrect
Narrow mode matches a primary element only against its own supporting elements; Wide mode uses all primary matches in range.
CUpload the table unhashed so EDM can compare spelling variants
Incorrect
Only hashed values of the source table are uploaded; the data is never in the open.
DMulti-token matching for HolderName
Correct
Multi-token matching lets EDM match corroborative values that contain several words, such as full names.
ESplit HolderName into two columns, because EDM can't match values that contain spaces
Incorrect
Multi-token matching handles values made of several words.
In the new experience you can select Ignore delimiters and punctuation for columns, and multi-token matching for corroborative fields whose values contain more than one word, such as names or addresses. Narrow mode matches a primary element only against its own supporting elements, while Wide mode uses supporting elements from all primary matches in range. The source table is hashed with a salt before upload.
You're writing a custom SIT in rule package XML for employee IDs. Each ID appears next to the employee's hire date, written in the format commonly used in the US, and the pattern should use that date as corroborating evidence. How can the pattern find the date without a regular expression of your own?
AUpload your own custom date function
BReference the built-in Func_us_date function
CMark the date field as searchable in an EDM schema
DSet the caseSensitive attribute on the date term
Show the answer and why
AUpload your own custom date function
Incorrect
Microsoft 365 doesn't support uploading your own custom functions; an entity can reference built-in functions only.
BReference the built-in Func_us_date function
Correct
A custom entity can use the built-in Func_us_date function to identify a date in the format commonly used in the US.
CMark the date field as searchable in an EDM schema
Incorrect
EDM schemas describe a source table; they don't add a date detector to this pattern.
DSet the caseSensitive attribute on the date term
Incorrect
caseSensitive makes keyword matches case-exact; it doesn't detect dates.
A pattern in a custom SIT can use functions that Microsoft 365 exposes. For example, an employee ID entity can use the built-in Func_us_date function to identify a date in the format commonly used in the US as corroborating evidence.
An analyst with access to Data explorer can preview most documents listed for the Credit card SIT, but one Word file in SharePoint shows no preview and can't be downloaded. What's the most likely reason?
AThe file is larger than the preview limit for Word files
BThe file is stored in SharePoint rather than in OneDrive
CThe file is encrypted
DThe file matched the SIT with low confidence
Show the answer and why
AThe file is larger than the preview limit for Word files
Incorrect
Data explorer doesn't block this file for its size; preview is turned off for another documented reason.
BThe file is stored in SharePoint rather than in OneDrive
Incorrect
Data explorer lists and previews items from both SharePoint and OneDrive.
CThe file is encrypted
Correct
Document preview and download are disabled for encrypted documents to mitigate security risks.
DThe file matched the SIT with low confidence
Incorrect
Confidence level doesn't decide whether a listed item can be previewed.
Data explorer shows where classified and labeled items are and lets authorized people review them. Document preview and download are disabled if the document is encrypted, to mitigate security risks.
Your organization turns on OCR scanning so that scanned contracts saved as images can be classified. Which classifiers can then detect characters in those images?
AOnly custom SITs created after OCR was turned on
BAll sensitive information types and trainable classifiers
COnly trainable classifiers, not sensitive information types
DOnly built-in SITs
Show the answer and why
AOnly custom SITs created after OCR was turned on
Incorrect
OCR isn't limited to SITs created after it's turned on.
BAll sensitive information types and trainable classifiers
Correct
When you enable OCR, all SITs and trainable classifiers can detect characters that are in images.
COnly trainable classifiers, not sensitive information types
Incorrect
SITs can detect characters in images too.
DOnly built-in SITs
Incorrect
Custom SITs and trainable classifiers can use OCR as well.
OCR scanning is an optional feature for scanning images for sensitive information. When you enable it, all sensitive information types and trainable classifiers can detect characters that are in images.