Development of an anonymity summary of VMS data ouptuts

[colinpmillar]

Rational:

• In order to guarantee that vessel anonymity is assured and transparent to data submitters and to the WGSFDGOV and ACOM, it is important to have a report summarising any anonymity issues for a given data ouput.
• Another benefit of such a report is to detail the steps taken to assure anonymity in the case that the standard outut does not conform. This will provide a caveats section for the data product being created.

Suggested sections:

• [x] Number of c-squares with vessels < 2 by category in the standard output
• [ ] effect of option 2.
• [ ] effect of option 4.

the following is taken from WGSFD 2019 https://doi.org/10.17895/ices.pub.5648:

• When data are being published, each unit should contain at least 3 vessels.
• Publication of data (including maps/charts/tables) must use one of the following techniques
  1. Suppression of data that includes less than 3 different vessels, by suppressing sensitive values.
  2. Aggregation of data so that each aggregation contains at least 3 different vessels.

Further: ICES WGSFD recommends that the following guidelines are followed when publishing data:

1. Swept area ratios (SAR) are not sensitive and can be published, even if there are less than 3 vessels within the aggregation. This information cannot be used to identify individual vessel.
2. If there is need to publish data with less than 3 vessels within the aggregation level, the data values can be classified, so that only groups are published (e.g. kw groups), that are wide enough that individual vessels can’t be identified.
3. Published data should not include information that can be used to infer the suppressed value (e.g. if the value of a single unit is suppressed but the total value is published then the suppressed value might be calculated).
4. A solution for publishing the sensitive data have been mentioned in the ICES data call that data are only made public at ICES rectangle level, but it would be possible to give information on the empirical distribution of values within each ICES rectangle.
5. Special requests for advice, for example the current work in support of WKBEDPRES2, can be addressed with data calls to national labs to produce rasters from point data which can then be aggregated at an international level.

[neil-ices-dk]

thanks @colinpmillar i have some doubts about the 'steps taken' as we should ensure we are taking the steps as outlined in the datacall/VMS licence and any other agreement - so this is 'standard' and we would therefore refer to these boiler plate texts - but do you mean to ensure a kind of checklist that these things have been adhered to?

[colinpmillar]

That's the lines I was thinking on. Best case scenario, the aggregation is high enough so we dont need to do anything, and we can report that, but in the case that we do have cells with anonymity issues, we list the standard steps that we had to take (referencing the WGSFD 2019 or other reference). I think it would be much safer if WGSFDGOV could review the procedure and report, then we can run this as a standard output for any VMS data request, maybe.

[colinpmillar]

@christianvdorrien suggests developing a draft of this for the December WGSFDGOV meeting using the data required for a previous Helcom request, so we can compare the output.

[colinpmillar]

draft report on anonymisation is here: https://github.com/ices-eg/wg_WGSFDGOV/blob/master/anonymity/report.html

and a preview can be seen here: https://htmlpreview.github.io/?https://github.com/ices-eg/wg_WGSFDGOV/blob/master/anonymity/report.html

infortunately it looks like the interactive plots are not showing, so I can demonstrate by screen sharing. I think if you download this file it will work when opened in a browser.

[colinpmillar]

• [ ] Apply this report to the Ospar area
• [ ] Develop the text of the report, to detail the method and list the options.
• [ ] Develop a list of issues - different spatial regions will give different values in anonymised cells.
• [ ] present to the WGSFD group by - Jan 2021
• [ ] and get comments from WGSFD by Feb 2021

Note the Ospar product should be published in Summer 2021

[colinpmillar]

Hi @ices-eg/wgsfdgov Just to check I am on the right track,

• I will develop the annonimity procedure to also include spatial grouping, and check the list proposed by WGSFD.
• write up the methodology and include in the report itself, so it is clear what was done - I can do a first draft, and then the WGSFD should take it and finalise. I will take the initial text from the WGSFD report
• Provide several options and a summary of the impacts on bias, spatial bias, etc.
• part of this process will be the creation of several alternative data products based on the different WGSFD proposals,
  • the first being that they only request SAR or kWh,
  • then spatal group,
  • value grouping (this is what I think people meant by rasterizing) accross the whole dataset, and
  • partial value grouping (grouping only applied to cells with < 3 vessels)

It would be good to produce a list of priorities / criteria, such as:

• spatial consistency over time
• Bias in terms of published value only
• Spatial bias in terms of spatial totals - we can make it so the sum over the whole region is unbiased, but different groupings will have different performance with respect to totals for a spatial subset.

Then we can summaries each procedure based on these priorities / criteria

Anyway, thoughts would be good, so just pop them in as a comments on this thread - cheers!

[jensr]

Hi @colinpmillar That sounds really good. I think one of the things I was getting a little confused about yesterday was whether the tool is for decision support only, or if you are looking to do it all in one go. One of the key things for WGSFDGOV to understand is probably the decision criteria, and to ensure these align with data policy/licenses/limitations in general, whereas WGSFD have the expertise to define what is useful and necessary.