How to early detect notification enabling (writting 1 to CCCD)?

[Alex Co]

Hi I would like to, at app level, implement access control to characteristic. My characteristic is configured solely as notify (no read, no writing).

When a smartphone turns on the notify feature of the characteristic I would like to decided if it can go one (and receive notifications) or deny it and disconnect (no permission to turn n notification).

Is it possible to implement that? How?

How can I early detect that a smartphone is just trying to write 1 to the CCCD?

Thanks a lot

Alex

[Edvin]

Hello,

The usual way to determine this is to restrict what devices that can change the CCCD before it happens, rather than when it has already happened. I don't think you can do that.

But the common way is to say that you need to be paired, or even use MITM protection in order to change the CCCD. This way, only known devices will be able to receive notifications. Would that work for you? If so, you can look into e.g. the way the ble_app_gls set up it's services and characteristics. 

Inside services_init() in main.c, you can see that e.g. the gls_init.gl_meas_cccd_wr_sec = SEC_JUST_WORKS. This means that the link has to be encrypted with "JUST_WORKS" to change the CCCD.

If it was set to SEC_OPEN, anyone can enable notifications, while if it was set to SEC_MITM, you would have to pair using MITM protection, such as an out of band passkey to be allowed to enable notifications.

[Alex Co]

Unfortunately, that seems not be useful for me as the device is to be mass produced and we can not forecast the devices that will request the notify feature...

As it is possible to intercept READ and WRITE I thought that it could be also possible to trap/intercept notify.

So, I have to implement that at app layer, right?

[Edvin]

May I ask then, how do you intend to decide what devices that can enable notifications?

[Alex Co]

If handled at app layer I can, for instance, require a password or token in order to turn on notify. Of course, notify will be enabled anyway but I can program it to send a error code 3 times in order to inform central device that password is not ok. If passord is ok I can notify the real values.

Just an idea...

[Edvin]

That is an idea that can work, but it is not very secure. If someone is sniffing this traffic, they will be perfectly capable of replicating these packets, so that your device thinks it is the correct device. 

I don't know how sensitive your data is, or if you are afraid of Denial of Service attacks (if someone does this all the time, you may not be able to connect with the correct central).

If this is not a concern of yours, then this approach should be fine. If you just want to prevent random phone apps trying to connect to your device.