Spirals-Team / FPStalker

Repo of code for FPStalker article
GNU Affero General Public License v3.0
31 stars 21 forks source link

Is it possible to track the length of an evolution in time within the provided dataset? #1

Open ErikThoreson opened 6 years ago

ErikThoreson commented 6 years ago

Hello,

Your research is interesting. As I play around with the example data in this repo I'm curious about a couple things. What do creationDate, updateDate and endDate mean?

Is it possible to track the length of an evolution in time before a break in the chain occurs?

Example query of the dt fields for a single user_tracking_id

select creationDate, updateDate, endDate from canvas_fp_project.extensionDataScheme where id='90c66a03-cbcf-4f9a-8330-e639509dbf9c' ;
+---------------------+---------------------+---------------------+
| creationDate        | updateDate          | endDate             |
+---------------------+---------------------+---------------------+
| 2015-11-28 05:00:00 | 2015-11-28 05:00:00 | 2015-11-28 05:00:00 |
| 2015-11-28 06:00:00 | 2015-11-30 10:00:00 | 2015-11-30 10:00:00 |
| 2015-11-30 23:00:00 | NULL                | 2015-11-30 23:00:00 |
| 2015-12-01 23:00:00 | NULL                | 2015-12-01 23:00:00 |
| 2015-12-01 23:00:00 | NULL                | 2015-12-01 23:00:00 |
| 2015-12-01 23:00:00 | 2015-12-01 23:00:00 | 2015-12-02 00:00:00 |
| 2015-12-02 00:00:00 | NULL                | 2015-12-02 00:00:00 |
| 2015-12-02 00:00:00 | 2015-12-03 13:00:00 | 2016-02-29 22:00:00 |
+---------------------+---------------------+---------------------+
antoinevastel commented 6 years ago

Hello Erik,

Amiunique stores fingerprints only when a change occurs in at least one of the attribute of the fingerprint. The creationDate corresponds to the time where the fingerprint was collected. updateDate corresponds to the last or first time (I'm not sure, I should verify in the code) where we verified if the fingerprint has changed, but no changed occurred. Finally, endDate corresponds to the date where the fingerprint changed. Warning: for privacy reasons, the times are rounded, so you may see two events appearing at the same time. Moreover, the database we distribute is only a sample of the whole database.

Concerning your second question -- "Is it possible to track the length of an evolution in time before a break in the chain occurs?" -- I'm not sure I understand it correctly. If you want to know how long a fingerprint, as characterised by its combination of attributes, remains constant, then you need to subtract endDate and creationDate.

I hope my explanations helped you, otherwise feel free to ask me any questions.

Antoine