docs: define sensitivity and protection method#151
Conversation
Greptile SummaryThis PR adds a new "Key concepts" section to
Confidence Score: 5/5Documentation-only change with no code modifications; safe to merge. The change adds a purely descriptive section to an existing doc file. The sensitivity weights and protection method names are consistent with the values already used in the leakage-mass formula and output-columns table elsewhere in the document, and they match the enum values in the codebase. No files require special attention. Important Files Changed
Flowchart%%{init: {'theme': 'neutral'}}%%
flowchart TD
A[Entity detected] --> B[Sensitivity assigned\nby disposition_analyzer]
B --> C{Sensitivity level}
C -->|high — weight 1.0| D[Direct identifier\nNames, IDs, contact details]
C -->|medium — weight 0.6| E[Quasi-identifier\nLocation, occupation, age]
C -->|low — weight 0.3| F[Generic attribute\nWidely shared traits]
D & E & F --> G[Protection method chosen\nholistic document view]
G --> H{Method}
H --> I[replace\nSynthetic alternative]
H --> J[generalize\nBroader form]
H --> K[suppress_inference\nRewrite surrounding text]
H --> L[remove\nDelete entity]
H --> M[leave_as_is\nNo change needed]
I & J & K & L & M --> N[Leakage scoring\nleakage_mass = Σ weight × confidence]
Reviews (5): Last reviewed commit: "fix: add more detail and organization to..." | Re-trigger Greptile |
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
| |-------|---------|---------|----------------| | ||
| | `high` | Exposure alone can identify a person | Names, ID numbers, contact details | 1.0 | | ||
| | `medium` | Meaningfully narrows the identity space | Location, occupation, age | 0.6 | | ||
| | `low` | Minimal standalone identifying power | Generic attributes, widely shared traits | 0.3 | |
There was a problem hiding this comment.
You could add gender here as an example as well!
| | `generalize` | Replaces the entity with a broader form | Quasi-identifiers (exact date → quarter, city → region) | | ||
| | `suppress_inference` | Rewrites the surrounding text to remove cues that enable the inference | Latent entities that are implied rather than stated | | ||
| | `remove` | Deletes the entity entirely | Cases where neither replacement nor generalization can preserve meaning without retaining the identifying detail | | ||
| | `leave_as_is` | Leaves the entity unchanged | Entities judged not to require protection in context | |
There was a problem hiding this comment.
Could also use gender as an example here
Changes include: