Secure e-mailing

CIZ Centre for Healthcare Indication

This algorithm helps e-mail users determine when an e-mail contains sensitive information and should be sent securely.

Last change on 2nd of December 2025, at 8:12 (CET) | Publication Standard 1.0

• Publication category: Other algorithms
• Impact assessment: DPIA
• Status: In use

General information

Theme

Organisation and business operations

Begin date

2025-12

Contact information

https://ciz.nl/hulp-en-contact#contact

Responsible use

Goal and impact

The purpose of this algorithm is to help ensure that information sent by e-mail is properly secured. In other words: not with too few security measures, but not with too many either. The algorithm contributes to this by alerting e-mail senders to sensitive information and encouraging them to then send this information securely. If they choose to do so, the e-mail is secured with encryption, two-factor authentication for the recipient, revocation option and logging.

This ensures that sensitive information is secured, while insensitive information can be accessed without extra effort for the recipient. The algorithm only impacts the process of sending e-mails and does not affect the substantive legal status of citizens or businesses.

Considerations

The use of this algorithm helps ensure that e-mails are sent with the right level of security. Classification of e-mails by algorithm is more effective than classification based on a glossary compiled by human input or classification by the sender, while at the same time the algorithm does not affect the content of an e-mail.

Human intervention

In principle, the user makes the choice whether or not to send an e-mail securely based on the recommendation by the algorithm. It is also possible to automate the choice of whether or not to send an e-mail securely based on the algorithm. In that case (not automatically sending insecurely), a user can still choose to explicitly send an e-mail securely.

Risk management

The overall performance of the algorithm is monitored by the vendor. If it appears that the algorithm is making incorrect classifications more frequently, this is picked up by the monitoring so that adjustments to the algorithm can be made. For the using organisation, the accuracy of Smart Classification is also visible in an administrator dashboard.

Legal basis

AVG

Links to legal bases

AVG: https://wetten.overheid.nl/BWBR0040940/2021-07-01

Impact assessment

Data Protection Impact Assessment (DPIA)

Operations

Data

The algorithm is based on anonymised historical messages combined with whether or not these messages were sent securely. Anonymisation is done in several steps, with only aggregates across many thousands of messages being stored. In addition, personal data is automatically removed from the data. With this, the data is anonymised and the original messages are impossible to extract from the data.

Technical design

The client's functioning and care needs are identified by the staff member during the indication examination. The findings of the examination are recorded by completing a questionnaire on impairment and (functional) limitations scoring. These questions are grouped into nine domains of self-reliance, and a scale score is calculated on each of these domains. The algorithm calculates from these scale scores, through weighting coefficients and cut-off values, which care profile fits best and which is the best alternative thereafter. Both are shown with their own percentage 'correctness'.

The weighting coefficients and cut-off values come from a regression analysis performed on historical data. There is no automatic adjustment of these values based on new data. Thus, the algorithm is not self-learning. This means that the CIZ can always explain on the basis of which logic the algorithm arrived at the advice.

External provider

Zivver