CoAP Servers + Border Routers on 6LoWPAN+RPL

Help archive
#1

[trying to re-send since the previous one got stuck at moderator’s review due to image attachment] Hello RIOTers,

I’m finally reaching to the mailing list to ask for some help, after a few days lurking around both users and devel mailling lists as well as Github notifications from the project.

I’ve seen a lot of keywords out there (e.g.: desvirt, ethos/ethos_br, tun/tap devices, ZEP, marz, tunslip, …), but I’m still unable to figure out the best way I could achieve the following scenario I want to build.

Just to provide context/motivation, and to present a few limitations as well, I would like to say:

  • I want to build this because it seems to cover, on a very general way, a typical scenario of distributed CoAP nodes forming a LoWPAN using the IETF stack (CoAP+UDP+RPL+6LoWPAN+802.15.4);
  • For my master thesis I want to investigate security scenarios/issues on IoT, and at least initially, I would like to do that with RIOT (great open source project, congratz guys!).
  • I believe this scenario would allow me to do this investigation over the whole stack of protocols, with nodes taking different roles and so on;
  • I would really like a way to accomplish this — or some closely-related form — using only native, because we’re very limited on re$ource$, but if needed we could try to buy some samr21-xpro (which seems to be the most developed in RIOT right now) or some cheaper hardware to use with XBee shields (or any other suggestion from you guys);

Basically, this is the scenario I have in mind: http://s13.postimg.org/8qmflmodj/scenario.png

​Is that possible/feasible? What would be the best (or at least least-problematic) way to achieve that given my thoughts above?

Thanks in advance, and sorry for the massive blob of a mail,

Bruno Melo.

Ps.: I was also able to run RIOT on Cooja, but as far as I understand I would have to develop a lot on Cooja to support the network interactions I would like to see (and also support a board to which I can compile CoAP server applications, since telosb doesn’t even have netif nor memory, from what I understood). This could be a way to go, but not my preferred one.

#2

Hi all,

Just to be sure, was this mail sent correctly/reached the list?

's, Bruno Melo.

#3

ping? :frowning:

Any help/directions would be appreciated.

#4

[putting you in CC because I don't know if you subscribed the mailing list]

Hi Bruno,

I'll give you a few hints:

I've seen a lot of keywords out there (e.g.: desvirt, ethos/ethos_br, tun/tap devices, ZEP, marz, tunslip, ...), but I'm still unable to figure out the best way I could achieve the following scenario I want to build.

desvirt [0] is the virtualization framework for the DES-Testbed (=actual hardware) [1] at FU Berlin. You can use desvirt to simulate arbitrary network topologies of RIOT nodes on your computer. The virtual nodes will be instances of the native port [2]. These will communicate using virtual network interfaces, so called tun/tap devices. A tun interface works on layer 3 (i.e. it gives you IP packets), a tap interface, which is what the native port uses, works on layer 2. tunslip [3] enables you to send IP packets over a serial interface. The border router makes use of this; the RIOT node sends IP packets over a serial interface to a Linux host, on which you can retrieve said packets from a tun interface. Concerning ZEP and marz, I'd say those are not important for you (I can't speak for ZEP, but marz was just a quick hack and to my knowledge doesn't work anymore).

Just to provide context/motivation, and to present a few limitations as well, I would like to say: - I want to build this because it seems to cover, on a very general way, a typical scenario of distributed CoAP nodes forming a LoWPAN using the IETF stack (CoAP+UDP+RPL+6LoWPAN+802.15.4);

RIOT supports 802.15.4, 6lowpan, IPv6, RPL, UDP, and CoAP. There's a tutorial for RPL in the wiki [4]. microcoap [5] is the CoAP library used by most RIOT folks, but I suggest you use this fork [6] instead, because microcoap is not being maintained anymore and lacks documentation.

- For my master thesis I want to investigate security scenarios/issues on IoT, and at least initially, I would like to do that with RIOT (great open source project, congratz guys!). - *I believe this scenario would allow me to do this investigation over the whole stack of protocols, with nodes taking different roles and so on*; - I would really like a way to accomplish this --- or some closely-related form --- using only native, because we're very *limited on re$ource$*, but if needed we could try to buy some samr21-xpro (which seems to be the most developed in RIOT right now) or some cheaper hardware to use with XBee shields (or any other suggestion from you guys);

Can you further specify what exactly you want to investigate? Basically, nothing in the IoT is secure in any way. You could start with the hardware that doesn't provide MMU's, operating systems (or the lack thereof) and their bugs, address spoofing, DDoS attacks, data integrity, encryption (DTLS) and so on.

​Is that possible/feasible? What would be the best (or at least least-problematic) way to achieve that given my thoughts above?

Again, depends on what exactly you want to look at.

Ps.: I was also able to run RIOT on Cooja, but as far as I understand I would have to develop a lot on Cooja to support the network interactions I would like to see (and also support a board to which I can compile CoAP server applications, since telosb doesn't even have netif nor memory, from what I understood). This could be a way to go, but not my preferred one.

I suggest you stick to RIOT OS on native and desvirt, then :wink:

Hope this helps; best regards and good luck, Lennart

[0] https://github.com/des-testbed/desvirt [1] http://des-testbed.net/node/4 [2] https://github.com/RIOT-OS/RIOT/wiki/Board%3A-native [3] https://github.com/RIOT-OS/RIOT/tree/master/dist/tools/tunslip [4] https://github.com/RIOT-OS/RIOT/wiki/Tutorial:-RIOT-and-Multi-Hop-Routing-with-RPL [5] https://github.com/1248/microcoap [6] https://github.com/i2ot/microcoap

#5

Hi Lennart,

Yes, I’m in the mailing list. First of all, thank you for the clarification regarding some of those tools. :slight_smile:

Until now I was able to create basic line topologies with desvirt, and also to follow the “RIOT and Multi-Hop Routing with RPL” successfully (besides the IOT-Lab part). I need to experiment with grid topologies, since I think that is more close to what I’m looking for.

My issue is that, although I’m able to run microcoap_server and gnrc_networking examples on native, the first uses IPv6+UDP+CoAP (so no RPL, 6LoWPAN and 802.15.4) and the second lacks 6LoWPAN and 802.15.4 (I’m ignoring CoAP in this case since I believe it would be “easy” to serve CoAP over that example).

It may be a somehow dumb questioning, but I wonder if I could run gnrc_networking (actually something based on it) on native including 802.15.4+6LoWPAN, although I don’t have the hardware — meaning some emulated form or anything close to this. I know 6LoWPAN is not necessary to the functionality of the example itself on native, but I’d like to see the packets/frames from those protocols going back and forth.

Can you further specify what exactly you want to investigate? Basically, nothing in the IoT is secure in any way. You could start with the hardware that doesn’t provide MMU’s, operating systems (or the lack thereof) and their bugs, address spoofing, DDoS attacks, data integrity, encryption (DTLS) and so on.

That’s the thing: we don’t know yet. I’m on a lab where people know their security-related stuff, but nothing about IoT tech. So, we wanted to build a “close enough to reality/production” scenario, and then start doing some Security Testing (from running static analysis tools to stress testing, some code coverage analysis), as well as looking for bugs/issues on the protocols specifications and [actually focus on] their implementations, trying to run DTLS, and so on.

Basically, we’re on a very exploratory phase. :frowning:

's, Bruno Melo.

Ps.: I’ll also change things here to use that microcoap fork, thank you for pointing me to that.

#6

Hi Bruno, following a discussion with @authmillenon, you might consider testing this PR: https://github.com/RIOT-OS/RIOT/pull/4861

Do you also have the impression that something like this would help you towards your goal? Cheers Emmanuel

#7

Oh you’re right, it really seems so. I’ve seen it before on GitHub but didn’t really noticed the potential.

I’ll take a look at this (PR+IETF Draft) and see what I can achieve with it.

Thanks a lot for the help, guys. I’ll keep in touch.

's, Bruno Melo.