search

merbanan / rtl_433

Program to decode radio transmissions from devices on the ISM bands (and other frequencies)
GNU General Public License v2.0
6.14k stars 1.33k forks source link

Add new decoder for Velux shutter remote control - io-homecontrol protocoll #1376

Open dwarning opened 4 years ago

dwarning commented 4 years ago

Recently I asked for differences in decoding Velux remote controller KI 313 signals with RTL_433 in flex decoder mode compared to manual decoding by the Pulseview tool, see https://groups.google.com/forum/#!topic/rtl_433/HxjH97Tkgys.

Attached is a zipped .cu8 file. Please look in the region from 540ms to 620ms for the four 7ms chunks. The second attachement is a screenshot from Pulseview with one data sequence decoded by the UART decoder 38.4k, 8N1 lsb first bitorder. In discussion with Karl Lohner the outcome was that there are some uncertainties regarding the start bit configuration in Pulseview and the capabilities of the built-in flex Decoder.

g001_868.91M_1024k.zip Pulseview_uart_decoded

The third attachement shows the decoding result for the Velux Controller in different operating modes like up, down and stop. by issuing the command two times to observe crc and rolling Code.

Velux_Codes

This is inline with a former work http://redd.it/cz8trb by z_rrr who sniffing the SPI between uC and RFIC in a Somfy remote which is also part of the io-homecontrol alliance.

I would like to ask for a new FSK_PCM Decoder, name it FSK_UART like Christian has suggested with the capability of specifing all needed decoding parameters, like Baudrate, protocol and start/stop bit configuration.

Don't hesitate to ask for more informations you need for implementation and testing.

Thank you, Dietmar

zuckschwerdt commented 4 years ago

Thanks for the research, I'll look into it soon. There are two possible ways to add support:

dwarning commented 4 years ago

Thank you, Christian. I want try to support you. I think the Problem is clear identified and could the solution for uart implementation anyway. Let me explain my way to decode. As I said I used the following conversion rtl_433 -s 1024k -w g001_868.91M_1024k.sr -r g001_868.91M_1024k.cu8 to get a Pulseview file. Then I converted the weak FM Signal to logic at first (and later we see wrong) with a conversion Level of 0.0V. So the very long synchronisation Phase (About 510ms, exact 1961 Bytes x55) starts wrong because of wrong start bit recognition. See picture Start_of_Sync_Phase Coming from that we are also false with start bit recognition in data phase: End_of_Sync_Phase Now I shifted the conversion level of Pulseview logic converter to -0.03V and got following Sync start phase: Start_of_Sync_low_level and the correct data stream with Velux specific start Bytes 0xFF33: End_of_Sync_low_conversion level So a common usable FSK_UART Decoder should take care for

I have no clue why the from IQ data calculated analog FM Signal is so weak (+/-55mV). I think to have stronger input for the pulseview conversion step would be helpful. Any comments are appreciated.

zuckschwerdt commented 4 years ago

Thanks for the very detailed analysis! The start- and stop-bit then have the same timing as every other bit. If we don't know about the Sync-word (start byte) then every bit could be seen as a start bit. I propose to do

  1. FSK_PCM decode
  2. search sync-word 0b0 0xff 0b1 0b0 0x33 0b1 (i.e. 0x7fd99, note that 0x33 is 0xcc "on the wire")
  3. run a decode_uart() function (a simple 10-to-8 decoder)

I guess we could try to find the start- and stop-bits by searching for a 0b10 pattern repeating after every 8 bits, but that seems like it could fail/false positive a lot.

I'll prototype something soon.

zuckschwerdt commented 4 years ago

UART decoder support has now been added. The syntax for the flex spec will likely change, but to test this right away use ,preamble=7fd99,decode_uart, e.g.: rtl_433 -c 0 -R 0 -X 'n=uart,m=FSK_PCM,s=26,l=26,r=7000,preamble=7fd99,decode_uart' -y 'ffff7fd9940481604' (not sure on the 26 µs, should match 38.4k?) Result should be:

codes     : {24}010203
dwarning commented 4 years ago

Made a executable with msys/mingw64 after git pull of master branch. If I use double quotes for flex options then it works with decode_uart on your test string with expected result.

So I used the velux remote control and got without decode_uart: `rtl_433 -f 868.89M -s 1024k -g 20 -c 0 -R 0 -X "n=uart,m=FSK_PCM,s=26,l=26,r=2048,preamble={16}0xaaff" rtl_433 version 20.02-49-g177fb6b branch master at 202005141540 inputs file rtl_tcp RTL-SDR Use -h for usage help and see https://triq.org/ for documentation.

New defaults active, use "-Y classic -s 250k" for the old defaults!

Disabling all device decoders. Registered 1 out of 152 device decoding protocols [ ] Found Rafael Micro R820T tuner [R82XX] PLL not locked! Sample rate set to 1024000 S/s. Tuner gain set to 20.700000 dB. Tuned to 868.890MHz.

time : 2020-05-14 17:45:13 model : uart count : 1 num_rows : 1 rows : len : 261 data : b326f80200802fc96eb395e00a028692e0080200ac2a49ea50862e6af299a5e768 codes : {261}b326f80200802fc96eb395e00a028692e0080200ac2a49ea50862e6af299a5e768

time : 2020-05-14 17:45:13 model : uart count : 1 num_rows : 1 rows : len : 262 data : b326f80200802fc96eb395e00a028692e0080200ac2a49ea50862e6af299a5e768 codes : {262}b326f80200802fc96eb395e00a028692e0080200ac2a49ea50862e6af299a5e768

time : 2020-05-14 17:45:13 model : uart count : 1 num_rows : 1 rows : len : 262 data : b326f80200802fc96eb395e00a028692e0080200ac2a49ea50862e6af299a5e76c codes : {262}b326f80200802fc96eb395e00a028692e0080200ac2a49ea50862e6af299a5e76c

time : 2020-05-14 17:45:13 model : uart count : 1 num_rows : 1 rows : len : 262 data : b326f80200802fc96eb395e00a028692e0080200ac2a49ea50862e6af299a5e76c codes : {262}b326f80200802fc96eb395e00a028692e0080200ac2a49ea50862e6af299a5e76c

time : 2020-05-14 17:45:14 model : uart count : 1 num_rows : 1 rows : len : 261 data : b326f80200802fc96eb395e00a028692e0080200ac264a625f886bfa72fbb46218 codes : {261}b326f80200802fc96eb395e00a028692e0080200ac264a625f886bfa72fbb46218

time : 2020-05-14 17:45:14 model : uart count : 1 num_rows : 1 rows : len : 262 data : b326f80200802fc96eb395e00a028692e0080200ac264a625f886bfa72fbb4621c codes : {262}b326f80200802fc96eb395e00a028692e0080200ac264a625f886bfa72fbb4621c

time : 2020-05-14 17:45:14 model : uart count : 1 num_rows : 1 rows : len : 261 data : b326f80200802fc96eb395e00a028692e0080200ac264a625f886bfa72fbb46218 codes : {261}b326f80200802fc96eb395e00a028692e0080200ac264a625f886bfa72fbb46218 But with decode_uart option there comes nothing: c:\msys64\home\dwarning\sdr_decode\velux\decode>rtl_433 -f 868.89M -s 1024k -g 20 -c 0 -R 0 -X "n=uart,m=FSK_PCM,s=26,l=26,r=2048,preamble={16}0xaaff,decode_uart" rtl_433 version 20.02-49-g177fb6b branch master at 202005141540 inputs file rtl_tcp RTL-SDR Use -h for usage help and see https://triq.org/ for documentation.

New defaults active, use "-Y classic -s 250k" for the old defaults!

Disabling all device decoders. Registered 1 out of 152 device decoding protocols [ ] Found Rafael Micro R820T tuner [R82XX] PLL not locked! Sample rate set to 1024000 S/s. Tuner gain set to 20.700000 dB. Tuned to 868.890MHz.

time : 2020-05-14 17:53:39 model : uart count : 1 num_rows : 1 rows : len : 0 data : codes : {0}

time : 2020-05-14 17:53:40 model : uart count : 1 num_rows : 1 rows : len : 0 data : codes : {0}

time : 2020-05-14 17:53:40 model : uart count : 1 num_rows : 1 rows : len : 0 data : codes : {0}

time : 2020-05-14 17:53:40 model : uart count : 1 num_rows : 1 rows : len : 0 data : codes : {0}

time : 2020-05-14 17:53:40 model : uart count : 1 num_rows : 1 rows : len : 0 data : codes : {0}

time : 2020-05-14 17:53:40 model : uart count : 1 num_rows : 1 rows : len : 0 data : codes : {0}

time : 2020-05-14 17:53:40 model : uart count : 1 num_rows : 1 rows : len : 0 data : codes : {0} ` Is there a misunderstanding of the preamble? Should I use the decoded value?

zuckschwerdt commented 4 years ago

If you have a PCM code of say aaffb326f80200..., i.e. after preamble matching b326f80200... then that won't work with UART. You are off by a single bit -- b32 is 1 01100110 0 10. See this BitBench with your codes.

You need to align one bit to the right, e.g. 55ff:

rtl_433 -c 0 -R 0 -X 'n=foo,m=FSK_PCM,s=26,l=26,r=2048,preamble=55ff,decode_uart' -y 'aaffb326f80200802fc96eb395e00a028692e0080200ac2a49ea50862e6af299a5e768'gives

codes     : {208}33f60000003fdacdea000161d20000000d255e0a18673d99e96e
dwarning commented 4 years ago

Yes - I can approve. preamble 55ff made the job:

time      : 2020-05-14 18:25:20 model     : uart         count     : 1             num_rows  : 1             rows      : len       : 208          data      : 33f60000003fdacdea000161d20000000d41e9d5c6c86a8c8b24 codes     : {208}33f60000003fdacdea000161d20000000d41e9d5c6c86a8c8b24

Some of the 8 telegrams have this bit shift we have discussed before. They come with 0xaaff and will be shown w/o preamble and decode option. But if I give preamble=aaff I got never the following data string.

zuckschwerdt commented 4 years ago

To clarify: the preamble can match anywhere in the data. No need to use all bits from the start. Preamble here means: search these bits and cut exactly there. You can also use the sync-word ff33 (encoded: 0 ff 1 0 cc 1 = 7fd99) to match:

rtl_433 -c 0 -R 0 -X 'n=uart,m=FSK_PCM,s=26,l=26,r=2048,preamble=7fd99,decode_uart' -y 'aaffb326f80200802fc96eb395e00a028692e0080200ac2a49ea50862e6af299a5e768' makes sure the full sync word is found and the output is only the payload:

codes     : {200}f60000003fdacdea000161d20000000d255e0a18673d99e96e
dwarning commented 4 years ago

Christian,

nevertheless I can't follow your encoding scheme below I can agree that with preamble 55ff (better 5555ff) all data can decoded from my remote control in right way.

Thanks for your efforts, Dietmar

psa-jforestier commented 4 years ago

Hello. I just realize I have Velux windows too. The IO-HOMECONTROL protocol is used by theses devices (I have the single-way remote control : no feedback from the window motor sent back to the remote). Thanks to @dwarning, I use the Flex decoder like this : rtl_433 -c 0 -R 0 -g 40 -X "n=uart,m=FSK_PCM,s=26,l=26,r=2048,preamble={24}0x5555ff,decode_uart" -f 868.89M

It gives me the following codes : Stop the motor : 

codes     : {208}33f60000003f708758000161d20000003bd2e6b62cef54c8a937
codes     : {208}33f60000003f708758000161d20000003bd6630743f0530ddc24

Up : 

codes     : {224}33f80000007f708758000161d40080c800003bd505526875499c7e72

Down : 

codes     : {224}33f80000007f708758000161d40080c800003bdf1ee7a0e304487a6b

Note : the down signal is detected 1 times on 3 or 4 tries.

The Command 1Byte you identified seems to be different, and also the messages have different length (224 for up and down, 208 for stop). I think the decoder is not accurate.

dwarning commented 4 years ago

Hello, don't know about the accuracy of this decoder. If I remember right I got also sometimes some bytes more then expected. Below you can find an excerpt of my results of investigation the velux protocoll.

In my opinion you have 27 bytes with Sync, Length, address, command, 7 bytes Rolling code and 2 bytes CRC-16 Kermit Polynom: 0x1021, Init 0x0000, Final XOR: 0x0000 at the end.

And the bytes for up, down and stop: `

F5 33 F6 20 00 00 3F DA CD EA 00 01 61 00 00 00 00 0B DD FD 8E F5 6F 15 AD AA 1E
F5 33 F6 00 00 00 3F DA CD EA 00 01 61 00 00 00 00 0B DD FD 8E F5 6F 15 AD 4F 9C
F5 33 F6 00 00 00 3F DA CD EA 00 01 61 00 00 00 00 0B DD FD 8E F5 6F 15 AD 4F 9C
D5 33 F6 00 00 00 3F DA CD EA 00 01 61 00 00 00 00 0B DD FD 8E F5 6F 15 AD 4F 9C
                                                     
FF 33 F6 20 00 00 3F DA CD EA 00 01 61 C8 00 00 00 0B BD 8A A3 A9 73 2E 10 26 D2
D5 33 F6 00 00 00 3F DA CD EA 00 01 61 C8 00 00 00 0B BD 8A A3 A9 73 2E 10 C3 50
FD 33 F6 00 00 00 3F DA CD EA 00 01 61 C8 00 00 00 0B BD 8A A3 A9 73 2E 10 C3 50
D5 33 F6 00 00 00 3F DA CD EA 00 01 61 C8 00 00 00 0B BD 8A A3 A9 73 2E 10 C3 50
                                                     
FX 33 F6 20 00 00 3F DA CD EA 00 01 61 D2 00 00 00 0B 99 15 DE CA CF 7E 0E 80 69
FF 33 F6 00 00 00 3F DA CD EA 00 01 61 D2 00 00 00 0B 99 15 DE CA CF 7E 0E 65 EB
FD 33 F6 00 00 00 3F DA CD EA 00 01 61 D2 00 00 00 0B 99 15 DE CA CF 7E 0E 65 EB
FF 33 F6 00 00 00 3F DA CD EA 00 01 61 D2 00 00 00 0B 99 15 DE CA CF 7E 0E 65 EB
                                                     
D5 33 F6 00 00 00 3F DA CD EA 00 01 61 D2 00 00 00 0B A1 05 17 5A 82 DF AE 8B BF
FF 33 F6 00 00 00 3F DA CD EA 00 01 61 D2 00 00 00 0B A1 05 17 5A 82 DF AE 8B BF
FF 33 F6 00 00 00 3F DA CD EA 00 01 61 D2 00 00 00 0B A1 05 17 5A 82 DF AE 8B BF
FD 33 F6 00 00 00 3F DA CD EA 00 01 61 D2 00 00 00 0B A1 05 17 5A 82 DF AE 8B BF

` Sorry - I can't attach the excel sheet. Saw no chance to recognize rolling code algorithm.

TheChatty commented 3 years ago

I tried psa's command and observed the following using my two Velux remotes (one code per button push):

UP 2xRIGHT, 2xLEFT, 1xPSA
codes     : {224}33f80000007fe1f573000161d40080c800000d6c2c3a3123e6ab7f1e
codes     : {224}33f80000007fe1f573000161d40080c800000d6ee448de7d4e0362d1
codes     : {224}33f80000007fc58967000161d40080c800000c6304e867ed64adf055
codes     : {224}33f80000007fc58967000161d40080c800000c658414991e8b06b82b
codes     : {224}33f80000007f708758000161d40080c800003bd505526875499c7e72

STOP 2xRIGHT, 2xLEFT, 2xPSA
codes     : {208}33f60000003fe1f573000161d20000000d6f708d89781e43bc24
codes     : {208}33f60000003fe1f573000161d20000000d71d1b10f26e1c18a9d
codes     : {208}33f60000003fc58967000161d20000000c664fcf56fb1c72d31e
codes     : {208}33f60000003fc58967000161d20000000c682025e049f331b64a
codes     : {208}33f60000003f708758000161d20000003bd2e6b62cef54c8a937
codes     : {208}33f60000003f708758000161d20000003bd6630743f0530ddc24

DOWN 2xRIGHT, 2xLEFT, 1xPSA
codes     : {224}33f80000007fe1f573000161d40080c800000d749fb9a0665ff477a6
codes     : {224}33f80000007fe1f573000161d40080c800000d7671b81065a2e20616
codes     : {224}33f80000007fc58967000161d40080c800000c6b56fcf691e6a92c74
codes     : {224}33f80000007fc58967000161d40080c800000c6ddaf0208646688fad
codes     : {224}33f80000007f708758000161d40080c800003bdf1ee7a0e304487a6b

I also got a rolling door using Somfy smoove io which produces roughly 10 codes per button push. E.g.

codes     : {208}33f60000003f17f523000147c800000018c438789cb680ccbc74
codes     : {200}33f60000003f17f523000147c800000018c438789cb680ccbc
codes     : {224}33f80000003f17f5232002ff0143010e000018c5045ee107363d59b4
codes     : {216}33f80000003f17f5232002ff0143010e000018c5045ee107363d59
codes     : {224}33f80000003f17f5232002ff0143010e000018c5045ee107363d59b4
codes     : {216}33f80000003f17f5232002ff0143010e000018c5045ee107363d59
codes     : {224}33f80000003f17f5232002ff01430105ff0018c6a34715cbe0124f7f
codes     : {224}33f80000003f17f5232002ff01430105ff0018c6a34715cbe0124f7f
codes     : {224}33f80000003f17f5232002ff01430105ff0018c6a34715cbe0124f7f
codes     : {224}33f80000003f17f5232002ff01430105ff0018c6a34715cbe0124f7f

Does this help?

zuckschwerdt commented 3 years ago

@dwarning did a very thorough analysis, and found a/the structure including CRC. We can check with your data and turn this into a decoder -- but there is still the rolling code encryption or MAC.

zuckschwerdt commented 3 years ago

From the previous examples and the new packets I assume to following format. The dest/src are not encrypted and the payload could be unencrypted too, i.e. it's only a MAC- not encryption-scheme.

33 f6 2000003f dacdea00 016100000000     0bdd fd8ef56f15ad aa1e
33 f6 0000003f dacdea00 016100000000     0bdd fd8ef56f15ad 4f9c
33 f6 2000003f dacdea00 0161c8000000     0bbd 8aa3a9732e10 26d2
33 f6 0000003f dacdea00 0161c8000000     0bbd 8aa3a9732e10 c350
33 f6 2000003f dacdea00 0161d2000000     0b99 15decacf7e0e 8069
33 f6 0000003f dacdea00 0161d2000000     0b99 15decacf7e0e 65eb
33 f6 0000003f dacdea00 0161d2000000     0ba1 05175a82dfae 8bbf

33 f8 0000007f e1f57300 0161d40080c80000 0d6c 2c3a3123e6ab 7f1e [UP RIGHT]
33 f8 0000007f e1f57300 0161d40080c80000 0d6e e448de7d4e03 62d1 [UP RIGHT]
33 f8 0000007f c5896700 0161d40080c80000 0c63 04e867ed64ad f055 [UP LEFT]
33 f8 0000007f c5896700 0161d40080c80000 0c65 8414991e8b06 b82b [UP LEFT]
33 f8 0000007f 70875800 0161d40080c80000 3bd5 05526875499c 7e72 [UP PSA]
33 f6 0000003f e1f57300 0161d2000000     0d6f 708d89781e43 bc24 [STOP RIGHT]
33 f6 0000003f e1f57300 0161d2000000     0d71 d1b10f26e1c1 8a9d [STOP RIGHT]
33 f6 0000003f c5896700 0161d2000000     0c66 4fcf56fb1c72 d31e [STOP LEFT]
33 f6 0000003f c5896700 0161d2000000     0c68 2025e049f331 b64a [STOP LEFT]
33 f6 0000003f 70875800 0161d2000000     3bd2 e6b62cef54c8 a937 [STOP PSA]
33 f6 0000003f 70875800 0161d2000000     3bd6 630743f0530d dc24 [STOP PSA]
33 f8 0000007f e1f57300 0161d40080c80000 0d74 9fb9a0665ff4 77a6 [DOWN RIGHT]
33 f8 0000007f e1f57300 0161d40080c80000 0d76 71b81065a2e2 0616 [DOWN RIGHT]
33 f8 0000007f c5896700 0161d40080c80000 0c6b 56fcf691e6a9 2c74 [DOWN LEFT]
33 f8 0000007f c5896700 0161d40080c80000 0c6d daf020864668 8fad [DOWN LEFT]
33 f8 0000007f 70875800 0161d40080c80000 3bdf 1ee7a0e30448 7a6b [DOWN PSA]

33 f6 0000003f 17f52300 0147c8000000     18c4 38789cb680cc bc74
33 f8 0000003f 17f52320 02ff0143010e0000 18c5 045ee107363d 59b4
33 f8 0000003f 17f52320 02ff01430105ff00 18c6 a34715cbe012 4f7f
^  ^  ^        ^        ^                ^    ^            ^CRC
^  ^  ^        ^        ^                ^    ^MAC
^  ^  ^        ^        ^                ^counter
^  ^  ^        ^        ^payload
^  ^  ^        ^source
^  ^  ^destination
^  ^length of payload
^sync, not included in CRC

It's interesting to note that at least the (presumed) destination, source, and payload field seem not included in the MAC (which is a security bug).

TheChatty commented 3 years ago

Is there a firmware file of any io homecontrol device available (to extract the signature private key)? --> There is.

PS: "RIGHT" and "LEFT" just refer to different sets of remote/window as well as "PSA" refers to psa's remote/window.

zuckschwerdt commented 3 years ago

It's a (truncated) CMAC, there is no indication of AES. And you really don't need it anyway.

zuckschwerdt commented 3 years ago

I've added support for Somfy io-homecontrol devices now. Please investigate all your devices and see if

time      : 2021-06-25 15:35:36
model     : Somfy-IOHC   id        : 00eacdda
Dest ID   : 3f000000     Msg type  : f6
Message   : 0161d2000000
Counter   : 3365         MAC       : 5e0a18673d99
Integrity : CRC
TheChatty commented 3 years ago

This might not be the whole message. If I compare up/down from the same window:

33 f8 0000007f e1f57300 0161d40080c80000 0d6c 2c3a3123e6ab 7f1e [UP RIGHT]
33 f8 0000007f e1f57300 0161d40080c80000 0d6e e448de7d4e03 62d1 [UP RIGHT]
33 f8 0000007f e1f57300 0161d40080c80000 0d74 9fb9a0665ff4 77a6 [DOWN RIGHT]
33 f8 0000007f e1f57300 0161d40080c80000 0d76 71b81065a2e2 0616 [DOWN RIGHT]
^  ^  ^        ^        ^                ^    ^            ^CRC
^  ^  ^        ^        ^                ^    ^MAC
^  ^  ^        ^        ^                ^counter
^  ^  ^        ^        ^payload
^  ^  ^        ^source
^  ^  ^destination
^  ^length of payload
^sync, not included in CRC

Thus the actual command (up or down) seems to be part of the "MAC".

And I'm interested in rebuilding the signature algorithm because I actually want to control the window (not receive commands).

zuckschwerdt commented 3 years ago

Yes, 6 bytes MAC does not feel right. It might be 2 or 4 byte data plus 4 or 2 byte actual MAC. Or it could just be encrypted bytes. But based on the whole scheme of this protcol I'd guess it's : fixed header + variable payload + checks (with checks being CMAC and CRC). Try to catch as many packets as possible, log preferably in JSON, and post a file here. Perhaps you can roll on of the senders counter up to match another one? That should be interesting. (PSA is way of, but LEFT can catch up to RIGHT in ~256 clicks...) Worst case log 65536 MACs and then table them :p -- it will also show if there is there is data in those.

TheChatty commented 3 years ago

Single button push usually increments counter by 2 thus I only had to push ~128 times ("down" btw). When counter was similar I produced a couple of same-counter-logs:

{"time" : "2021-06-25 17:49:07", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fc58967000161d40080c800000d9a87ac970b53d1d80d"}], "codes" : ["{224}33f80000007fc58967000161d40080c800000d9a87ac970b53d1d80d"]}
{"time" : "2021-06-25 17:50:37", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fe1f573000161d40080c800000d9a4c0e4a0e45aa995d"}], "codes" : ["{224}33f80000007fe1f573000161d40080c800000d9a4c0e4a0e45aa995d"]}
{"time" : "2021-06-25 17:49:08", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fc58967000161d40080c800000d9cf610e5e0eb9f6ba1"}], "codes" : ["{224}33f80000007fc58967000161d40080c800000d9cf610e5e0eb9f6ba1"]}
{"time" : "2021-06-25 17:50:39", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fe1f573000161d40080c800000d9cec44bf478a060cad"}], "codes" : ["{224}33f80000007fe1f573000161d40080c800000d9cec44bf478a060cad"]}
{"time" : "2021-06-25 17:49:11", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fc58967000161d40080c800000d9e10707b9a1f4aed84"}], "codes" : ["{224}33f80000007fc58967000161d40080c800000d9e10707b9a1f4aed84"]}
{"time" : "2021-06-25 17:50:41", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fe1f573000161d40080c800000d9ede2cc3c1117af8fd"}], "codes" : ["{224}33f80000007fe1f573000161d40080c800000d9ede2cc3c1117af8fd"]}
{"time" : "2021-06-25 17:49:13", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fc58967000161d40080c800000da011280dad6fe7b06a"}], "codes" : ["{224}33f80000007fc58967000161d40080c800000da011280dad6fe7b06a"]}
{"time" : "2021-06-25 17:50:43", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fe1f573000161d40080c800000da0209550b7e3530a35"}], "codes" : ["{224}33f80000007fe1f573000161d40080c800000da0209550b7e3530a35"]}
{"time" : "2021-06-25 17:49:15", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fc58967000161d40080c800000da2445f4c017d240cde"}], "codes" : ["{224}33f80000007fc58967000161d40080c800000da2445f4c017d240cde"]}
{"time" : "2021-06-25 17:50:46", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fe1f573000161d40080c800000da2877cc2f46a388a5c"}], "codes" : ["{224}33f80000007fe1f573000161d40080c800000da2877cc2f46a388a5c"]}
{"time" : "2021-06-25 17:49:16", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fc58967000161d40080c800000da4787e20402f196b43"}], "codes" : ["{224}33f80000007fc58967000161d40080c800000da4787e20402f196b43"]}
{"time" : "2021-06-25 17:50:47", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fe1f573000161d40080c800000da44412916d3fe642b8"}], "codes" : ["{224}33f80000007fe1f573000161d40080c800000da44412916d3fe642b8"]}
{"time" : "2021-06-25 17:49:17", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fc58967000161d40080c800000da633a9c2c0cc665e5d"}], "codes" : ["{224}33f80000007fc58967000161d40080c800000da633a9c2c0cc665e5d"]}
{"time" : "2021-06-25 17:50:49", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fe1f573000161d40080c800000da60eb34055b18f1209"}], "codes" : ["{224}33f80000007fe1f573000161d40080c800000da60eb34055b18f1209"]}
{"time" : "2021-06-25 17:49:19", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fc58967000161d40080c800000da8b3f7dd7a366fde5e"}], "codes" : ["{224}33f80000007fc58967000161d40080c800000da8b3f7dd7a366fde5e"]}
{"time" : "2021-06-25 17:50:51", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fe1f573000161d40080c800000da807c9a103761b1f5b"}], "codes" : ["{224}33f80000007fe1f573000161d40080c800000da807c9a103761b1f5b"]}
{"time" : "2021-06-25 17:49:20", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fc58967000161d40080c800000daa80075cf2e7bc6064"}], "codes" : ["{224}33f80000007fc58967000161d40080c800000daa80075cf2e7bc6064"]}
{"time" : "2021-06-25 17:50:53", "model" : "uart", "count" : 1, "num_rows" : 1, "rows" : [{"len" : 224, "data" : "33f80000007fe1f573000161d40080c800000daaeb6adf4f8fad3404"}], "codes" : ["{224}33f80000007fe1f573000161d40080c800000daaeb6adf4f8fad3404"]}
zuckschwerdt commented 3 years ago

Great! You are right then, there is data in the encrypted field, not just a MAC over the counter.

f8 0000007f e1f57300 0161d40080c80000 0d9a 4c0e4a0e45aa 995d
f8 0000007f c5896700 0161d40080c80000 0d9a 87ac970b53d1 d80d
f8 0000007f e1f57300 0161d40080c80000 0d9c ec44bf478a06 0cad
f8 0000007f c5896700 0161d40080c80000 0d9c f610e5e0eb9f 6ba1
f8 0000007f c5896700 0161d40080c80000 0d9e 10707b9a1f4a ed84
f8 0000007f e1f57300 0161d40080c80000 0d9e de2cc3c1117a f8fd
f8 0000007f c5896700 0161d40080c80000 0da0 11280dad6fe7 b06a
f8 0000007f e1f57300 0161d40080c80000 0da0 209550b7e353 0a35
f8 0000007f c5896700 0161d40080c80000 0da2 445f4c017d24 0cde
f8 0000007f e1f57300 0161d40080c80000 0da2 877cc2f46a38 8a5c
f8 0000007f e1f57300 0161d40080c80000 0da4 4412916d3fe6 42b8
f8 0000007f c5896700 0161d40080c80000 0da4 787e20402f19 6b43
f8 0000007f e1f57300 0161d40080c80000 0da6 0eb34055b18f 1209
f8 0000007f c5896700 0161d40080c80000 0da6 33a9c2c0cc66 5e5d
f8 0000007f e1f57300 0161d40080c80000 0da8 07c9a103761b 1f5b
f8 0000007f c5896700 0161d40080c80000 0da8 b3f7dd7a366f de5e
f8 0000007f c5896700 0161d40080c80000 0daa 80075cf2e7bc 6064
f8 0000007f e1f57300 0161d40080c80000 0daa eb6adf4f8fad 3404
TheChatty commented 3 years ago

Thus I'd need a log of 32k pushes (+2/push) of up, 32k pushes of down and 32k pushes of stop be able to control the blinds. And then again for the other window ;-)

dwarning commented 3 years ago

Hello again, these guy comes to another conclusion https://community.openhab.org/t/io-homecontrol-velux-somethings-in-the-bush/11413/223 And there are other hints in web that io-homecontrol is using AES.

zuckschwerdt commented 3 years ago

Great find. AES is 128-bit-block based, so there is padding and truncation. The fixed IV is not a bug or problem -- you'd need to transmit a variable IV anyway, otherwise.

The CRC-16 is not 0x8408 but the standard 0x1021 but with bit reflection -- note that (implied)top and bottom bit are always set for a proper CRC.

zuckschwerdt commented 3 years ago

At a closer look the linked docs are for a wired communication, not radio. And since we only have 6 bytes of encryption it's either just a truncated CMAC or some encryption scheme other than AES. It could be as simple as a 16-bit LFSR generating an XOR key.

Hofyyy commented 3 years ago

Since Rojaflex is running. Now the Somfy "Situs 5 io" is the next step.

Edit: See logfiles attached from Situo5 io Pure II. Channel 3 up and down. somfy_io_situo5_ch3_down.txt somfy_io_situo5_ch3_up.txt

zuckschwerdt commented 3 years ago

If you see it on 868.89 then tune a little off e.g. 868.7 Are you able to get raw data? This should work rtl_433 -c 0 -R 0 -X "n=uart,m=FSK_PCM,s=26,l=26,r=300,preamble=5555ff,decode_uart" -f 868.7M

The logs look like noise. You never want to see pulses or gaps that short (< 10 µs).

Hofyyy commented 3 years ago

I do not understand why you did know that I have to go to 868.7 but it worked out of the box! 868.7 and 868.8 are working. You will find my values attached.

What is the current challenge the crack the whole protocol?

Do we need the 32000 up downs or are the information here enough? https://community.openhab.org/t/io-homecontrol-velux-somethings-in-the-bush/11413/223

If we would need that I could create a Raspberry connected to a remote and do that in an automated way ...

somfy_io_situo5_ch3_down.txt somfy_io_situo5_ch3_up.txt

zuckschwerdt commented 3 years ago

Summary on Somfy io-homecontrol:

This means:

TheChatty commented 3 years ago

I posted logs of same action and counter but different remote/window here.

merbanan commented 3 years ago

This might be the previous somfy protocol:

https://pushstack.wordpress.com/somfy-rts-protocol/

Maybe it is simlar to this protocol.

Hofyyy commented 3 years ago

I had a look into the Source and Destination IDs.

  1. Remote is sending first to Motor 1 (CH1) and then to Motor 2 (CH2)

See Motor 1 and Motor 2 communication and Source IDs. After that I testet CH3, CH4 and Group on the same remote without a Motor.

I could see that the source id is counting up in conjunction to the channels.

The I can see that the source ID changes e.g. from 0033bd2a -> 2033bd2a regarding to the dest id?!?! 0033bd2a for DEST id: bf010000 2033bd2a for DEST id: 3f000000

motor1.txt motor2.txt SourceID.txt

merbanan commented 3 years ago

https://service.somfy.com/downloads/nam_v5/somfyio_homecontrol_guide.pdf

"Encrypted Dialogue is 100% secure: The tamperproof 128-bit encryption key will thwart any hacking attempt. This encryption key is exclusive to each installation and provides the same level of security as an online bank transaction."

merbanan commented 3 years ago

Page 21, AES 128 bit.

merbanan commented 3 years ago

Encrypted emissions: the key to security

Each io-homecontrol® installation has its own encryption key, which is present in all io products in the home. This key is automatically activated when the installation is first used. The emitter (the remote control) issues its encryption key to the receiver (e.g. an io roller shutter) once and once only.

How does it work? • For each command issued by the remote control, the receiver generates and sends back to the emitter a random number generated from a range of several billions. • The emitter and receiver both perform automatic calculations based on this random number and the encryption key. • If the results of these two calculations are identical, the emitter and receiver must have the same key, and the command (e.g. close the shutter) can therefore be carried out (e.g. close the shutter). The emitter is then informed that the command has been carried out. The encryption key is “buried“ among these exchanges between emitters and receivers, making it undetectable.

zuckschwerdt commented 3 years ago

So these 6 bytes are a CMAC (truncated result of AES over some key+data). CMAC implies no hidden data in the encrypted field, which is strange. The payload field seems to be the same for different actions?

zuckschwerdt commented 3 years ago

Great list. I got the IDs wrong then. Seems 2abd or possibly f2abd3 is th ID. No clue what the 01bf / 003f field means. The 2/3/4/5/6 near then end of the "header" are channel/group like you have shown. The 00/20 byte before the "payload" seems to match the payloads 01/02 start.

Situo 5 io Pure II

1. Source ID

CH1: 0x0032 bd2a
Ch2: 0x0033 bd2a
CH3: 0x0034 bd2a
CH4: 0x0035 bd2a
Group CH1;2;3;4: 0x0036 bd2a

Motor CH 1 - Runter
{216} f8 0000 01bf 2abd 3200  0143 c800 80d3 0000  01d2 fc84b5650d24 1656

Motor CH 1 - Stop
{200} f6 0000 003f 2abd 3200  0143 d200 0000       01d7 5ef3118c15a2 e21a
{216} f8 0000 003f 2abd 3220  02ff 0143 0205 ff00  01d9 6fae22d98323 4739

Motor CH 1 - Hoch (rein)
{216} f8 0000 01bf 2abd 3200  0143 0000 80d3 0000  01da 1eadbc35da9b 573c
{216} f8 0000 003f 2abd 3220  02ff 0143 000c 0000  01dc 37b4b588adba b94b
{216} f8 0000 003f 2abd 3220  02ff 0143 0005 ff00  01dd 657b181ba62b 5a5a
{216} f8 0000 01bf 2abd 3200  0143 0000 80c8 0000  01de 382867e4d013 6849

Motor CH 2 - Runter (raus) (Down)
{216} f8 0000 01bf 2abd 3300  0143 c800 80d3 0000  010c 6d6075c82240 65d7
{216} f8 0000 003f 2abd 3320  02ff 0143 010c 0000  010e bb37771da469 c8d7
{216} f8 0000 003f 2abd 3320  02ff 0143 0105 ff00  010f f800ee67d4ae 48cd
{216} f8 0000 01bf 2abd 3300  0143 c800 80c8 0000  0110 8092c040c87a 5d91

Motor Ch2 - Stop
{200} f6 0000 003f 2abd 3300  0143 d200 0000       0111 97af97e98bca 719b
{216} f8 0000 003f 2abd 3320  02ff 0143 0205 ff00  0113 55ad8d2dc233 502a

Motor CH2 - Hoch (Up)
{216} f8 0000 01bf 2abd 3300  0143 0000 80d3 0000  0114 b66199af0306 1f42
{216} f8 0000 003f 2abd 3320  02ff 0143 000c 0000  0116 7207733698c5 7ce7
{216} f8 0000 003f 2abd 3320  02ff 0143 0005 ff00  0117 eb87d5a5b3bd feee
{216} f8 0000 01bf 2abd 3300  0143 0000 80c8 0000  0118 d346d3cadb75 ac25

The good news: the payload does actually encode the command. It's multiple different payload per command. But distinct per command. In the payload 80d3 could be the "button-pressed", 80c8 the "button-released", and the 2fff lines "button-still-pressed" with a counter in that third word.

Also the payload might start one byte earlier, the CMAC-counter would then be 24-bit (which fit the billions number better). E.g.

{216} f8 0000 01bf 2abd 33  0001 4300 0080 d300  000114 b66199af0306 1f42
{216} f8 0000 003f 2abd 33  2002 ff01 4300 0c00  000116 7207733698c5 7ce7

We'd known when we observe the counter going over ffff ;)

merbanan commented 3 years ago

The description in this pdf is probably pretty close to the actual design. But logic dictates that before encryption can occur both parties must have the encryption key.

"The emitter (the remote control) issues its encryption key to the receiver (e.g. an io roller shutter) once and once only."

So this can happen in plain text or be scrambled somehow. Imo scrambling is the only option ie a shared secret. And with the protocol being backed by an association this is how you share the secret.

Anyway this is probably quite complicated. The fastest path forward is probably to reverse engineer a bridge or something like that.

merbanan commented 3 years ago

The protocol is also bi-directional if I understand the document text correctly. So the logs might be from the remote sending and the receiver responding. And it implied that this is done for the remote to get feedback that the action is performed properly.

merbanan commented 3 years ago

There exist a bridge called KLF 200. It is a io-homecontrol gateway. I am not 100% sure it is a radio compatible device but it has a documented api:

https://velcdn.azureedge.net/-/media/com/api/klf200/technical%20specification%20for%20klf%20200%20api-ver3-18.pdf

Page 26-27 might describe the structure of the ids. If the device talks the same or a similar protocol then more things can probably be mapped to the observed rf-payload.

merbanan commented 3 years ago

@Hofyyy can you log some messages with -m level active? I think we are seeing transmissions from 2 different devices. Keep close to one of the devices.

Hofyyy commented 3 years ago

Distance:

See new log files motor1.txt motor2.txt

zuckschwerdt commented 3 years ago

@TheChatty can you record a STOP for each of RIGHT, LEFT, PSA and each channel: 1,2,3,4,group? 3x5 =15 messages.

merbanan commented 3 years ago

If I look at the motor1.txt we can see 4 counter values. 487, 489, 490 and 491.

487 has 34.2 dB SNR (assumed to be remote)        0143c80080d30000  2efe3e9ff7a6
489 has 32.0 dB SNR (assumed to be motor ch1)     02ff0143010c0000  8b36f538266b
490 has 34.2 dB SNR (assumed to be remote)        02ff01430105ff00  dc46552cc9cc
491 has 34.2 dB SNR (assumed to be remote)        0143c80080c80000  67481214aafb

Based on the pdf document about io-homecontrol a possible logic flow could be the following:

487 - Hello. Execute down command (remote) 489 - Ok, I have verified the authenticity of the command (receiver) 490 - Ok, I have verified the authenticity of your response command (remote) 491 - Good bye (remote)

So IMO we can clearly see 2 devices communicate here. The above flow is a guess but the pattern is visible in other logs.

zuckschwerdt commented 3 years ago

Should be best to record a sample file with the full conversation. rtl_433 -f 868.7M -w file_868.7M_1024k.cu8 -T 10 The quickly trigger the remote. Upload as zip.

merbanan commented 3 years ago

Yeah, maybe we are missing one signal. The counter jumps 2 for some reason. Maybe we are missing a packet.

Hofyyy commented 3 years ago

I see some messages which we are not interested in. time : 2021-06-27 19:40:28 model : Marlec-Solar Raw data : 42ee....

Anyway ... see logs attached. "Down" -> "Stop" -> "Up" CH1_868.7M_1024k.cu8.zip CH2_868.7M_1024k.cu8.zip

merbanan commented 3 years ago

Ok, this signal is pushing what the code is able to handle.

The messages are repeated 3 or 4 times.

It looks like we are missing one message indicated by the counter because there is a long preamble filling up the buffer.

The framing code fails in some cases also but the repeats will most likely recover the message because of the repeats.

merbanan commented 3 years ago

Down results in 5 messages transmitted. Stop gives 3 messages. Up gives 5 messages

Using the SNR values was a red herring. You need to set the gain fixed also. With dynamic gain things will be all over the place.

The logs we have seen so far is missing the initial reply.

merbanan commented 3 years ago

Anyway I looked into the actual hardware for this protocol and they are using the SiLabs EFM32. Several devices implementing the protocol seems to have this chip. This is based on information found here:

https://groups.google.com/g/openhab/c/AinJdyyDtG0

and here:

https://community.openhab.org/t/io-homecontrol-velux-somethings-in-the-bush/11413

From the info there it seems like this protocol is fairly secured. The mcu has built storage and and hardware aes engine:

https://i1.wp.com/limitedresults.com/wp-content/uploads/2021/06/arch_block.png?resize=1024%2C463&is-pending-load=1#038;ssl=1

All this together could make this system very secure because all the secrets could be handled inside the mcu. We most likely need this code to be able to figure out the protocol completely.

So we need to replicate this:

https://limitedresults.com/2021/06/enter-the-efm32-gecko/