nRF Sniffer decryption is not working after entering passkey

[Acutetech]

I am a newbie with nRF Sniffer, which I have installed on one PCA10040 board. I am sniffing communications between a Samsung phone running nRF Connect and a second PCA10040 running the glucose sample app.

I believe I am following the instructions in the User Guide section 5.5: when the glucose app prints the 6-digit passkey I enter this into Wireshark and type enter, then I enter it into the phone. The phone and peripheral then bond and work, but Wireshark starts printing "Encrypted packet decrypted incorrectly (Bad MIC)" . When I click on Log I see this:

INFO: Setting Passkey: 691295

INFO: Sent key value to sniffer: [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 10, 140, 95]

A screenshot is here:

/resized-image/__size/640x480/__key/support-attachments/beef5d1b77644c448dabff31668f3a47-e588f28969f94ee0a39b08c9e8afde0d/Screenshot-2018_2D00_11_2D00_19-10.57.32.png

Q1: Are my results expected? What should I do differently?

Q2: Assuming I am able to get beyond this problem, is it possible to sniff subsequent connections? The Central and Peripheral will be bonded and the peripheral will not print a passkey. Can nRF Sniffer decrypt the second session based on keys it has saved?

[David Edwin]

Can you mention the wireshark version used, nRF sniffer version used and board version.

Q2: Yes, nRF Sniffer should be able to sniff the subsequent re-connections.

Q1: Seems to be as expected on actions. On the timing of actions, ensure that the wireshark has the PIN before you tap to accept on the device screen (which you seem to have done as well).

Ensure you are using the nRF Sniffer v2 Beta 3 .

I will investigate after you have posted the versions of the software and hardware as requested above.

[Acutetech]

I installed software last Friday, so should be "latest".

Wireshark says 2.6.4

I have two nRF52 boards with labels that say PCA10040 1.2.4 2018.39 (nRF sniffer) and 2018.49 (Peripheral).

nRF Sniffer was extracted from a zip file nrf_sniffer_2.0.0-beta3_12oct2018_1c2a221.zip

I confirm that I entered the passkey in Wireshark before entering it in the Central.

The peripheral, slave and central are within say 75cm of each other. AFAIK I disable BLE on other nearby devices.

I have just tried sniffing a different peripheral which uses Just Works bonding (no passkey). I continued to use nRF Connect on the Samsung phone. They bond correctly but nRF Sniffer behaves the same, with "Encrypted packet decrypted incorrectly" immediately after the LL_START_ENC_REQ packet.

[Acutetech]

OK David - I put ble_app_cli in one PCA10040 and connect to it from nRF Connect running on Windows. I then bonded (clicked "Pair", checked only the "Perform Bonding" check box, selected "No keyboard/no display") Both nRF Connect and the peripheral report that they bonded, but nRF sniffer prints "decrypted incorrectly" again. Screenshot and sniffer log attached.ble_app_cli.pcapng

I don't see how I can delete the bonding information on both deveices so I can try again...

[David Edwin]

Erase and Reflash the board running the peripheral to delete the bond and try again. 

For nRF connect , click on the adapter wheel, then click "delete bond information".

I can see that the encrypted packets are being decrypted as "Encryption Information" and "Master Identification" are all encrypted packets.

In the Nordic_CLI device, click on the UART over BLE and then click UART TX. Click on the arrow so that the Nordic CLI will send periodic notifications.

It is possible some packets are not decrypted correctly if they are received incorrectly on the sniffer, but that will not impact the entire connection.

Read the user guide to setup the wireshark profile for the sniffer. (section 2.3.2 and section 2.3.3)

[Acutetech]

I have tried this cycle several times and almost always the decryption fails. I am about to give up with a "broken for me" conclusion. I did get one session in which I successfully bonded and then nRF Sniffer/Wireshark could decrypt each of the (encrypted) notifications from the peripheral, but almost always I get the "packet decrypted incorrectly" messages.

Interestingly, my most recent session (attached) shows bonding at around packet 4401, and some packets correctly decrypted until packet 4444, then failed packets with a few correctly decrypted packets interspersed (e.g. 5540, 5672, 5686). From 4444, the only successfully decrypted packets are sent by the slave - I don't know whether this is significant. Could there be some timing issue? e.g. packets come too fast to be processed?

Probably an unrelated bug/feature: whenever I ask nRF Connect to reconnect to the bonded peripheral I get a "failed to authenticate" dialog, even though the nRF Connect behaviour shows that it has connected correctly. 

ble_app_cli_2.pcapng

[David Edwin]

If you want to re-connect to bonded peripheral, then the "perform bonding" check box must be ticked when in the pairing screen.

Keep the boards close by for the test, maybe a few centimeters apart, with the sniffer in the center.

[Acutetech]

No better. The boards are almost touching. Most (but not all) packets give "packet decrypted incorrectly" messages. There is no sign of packets being lost or corrupted prior to encryption being enabled.

Have you actually tried the identical setup yourself?

[David Edwin]

sniffer_pca10040_mic_recover_v2_18Dec2018.hex

This seems to perform better for MIC failure. can you test this and let me know.

Edit: Updated hex fix

[bill.diken]

Acutetech I'm having an identical issue with Wireshark 3.0.1 with the latest nrf sniffer nrfsniffer200beta312oct20181c2a221 with the nRF52840 DK board. I can't seem to get decryption to work reliably, I do sometimes see correct value at the attribute even with the MIC error message but it only works for the first few transmissions before it't completely broken. Were you ever able to resolve this issue?

[Acutetech]

I could not resolve the problem and gave up.

This is pure speculation: I had another problem with notification data being corrupted, which was later fixed by a library update, or similar (I think to BLE Library 2.0.4). I wonder if there could be a connection? For this other problem see devzone.nordicsemi.com/.../167037

[David Edwin]

There is a new release being worked on to fix encryption issues and seems to perform better.nrf_sniffer_2.0.0-temp_beta-4_2May2019_47b3f5f.zip

You can try this and let me know if this works better. Unfortunately this snapshot that I uploaded here has not yet been tested for passkey but has been tested for Justworks pairing.

Edited: The sniffing of encrypted link  is improved only for the nRF52xxx based development kits, the performance of the nRF51 based development kits when used as sniffer remains the same as the previous release.

[bill.diken]

David, I used the file you provided, although I had to change the contents of the nrf_sniffer.bat file to get Wireshark to recognize the devboard.

@echo off
python "%~dp0nrf_sniffer.py"  %*

The bad MIC error is still there, it seems that it decrypts the first few transactions correctly and then starts to fail and doesn't recover. What seems to work better is the disconnect/reconnect detection, in the previous build I could never get packets when there was a disconnect and the strap started advertising or when it tried to establish a new connection.