Technical Tip: DLP regex to block words with accented characters does not work on ARM64-based platforms

[GGMACHAIN]

Description	This article describes how to solve the problem of DLP regex blocking words with accented characters on entry-level (ARM) FortiGate models (e.g.  FGT100F) in FortiOS 7.2.8. The steps below are the correct procedure for blocking words such as 'pepitá' using regular expressions via DLP FortiGate.
Scope	FortiGate.
Solution	In this example, the word 'pepitá' will be blocked. Create a new 'Dictionary' in Security Profile -> Data Leak Prevention -> Dictionaries and select 'Create New'. Create a new 'Sensor' in Security Profile -> Data Leak Prevention -> Sensor and select 'Create New'. Create a DLP Profile using the 'Sensor' profile created in step 2 with action 'Block', Type 'Message', and protocol 'HTTP-POST'. Important note: Use the DLP profile and policy in 'Proxy' mode and also enable 'deep-inspection' in the firewall policy. Via CLI change DLP profile to 'set feature-set proxy' The steps above are correct for blocking the example word 'pepitá', but do not work for equipment on entry-level (ARM) running FortiOS 7.2.x. This issue has been solved in 7.4.5 and 7.6.1. There is no permanent fix for 7.2. Workaround: It is necessary to adjust it by changing the Pattern in the Dictionary from '/pepitá/i' to only 'pepitá'. There are several websites for word tests on DLP systems: Example: https://dlptest.com/http-post/ This is an example of how FortiGate should behave and what message is displayed: