Simulating Security in Wireless Sensor Network

[vinay vishwas]

Hello ,
          Is NS3 suitable for simulating security aspect for Wireless Sensor Networks. To be specific , can I simulate the situation in which I can store some set of keys to sensor nodes and , nodes will communicate with each other to exchange this key information using handshaking. Then they will use keys to communicate with each other . i mean encrypt and decrypt with those keys.... 


VINAY VISHWAS

[Konstantinos]

Hi,

There are certain aspects you are asking that are not implemented in ns-3 and you might need to make certain assumptions.

For example, there is no security module in ns-3 to provide encryption/decryption methods. However, you should be able to use Crypto++ or any other cryptographic library.

Also, the communication module for sensor networks is not yet merged with ns-3 main tree. You can have a look here http://www.nsnam.org/wiki/Lr-wpan

You can build a wireless scenario (with wifi or lr-wpan patch) and create your own applications (client/server) to initially send unencrypted messages and then only encrypted.

Regards,

K.

[Tommaso Pecorella]

Hi,

yes, no, maybe.

The point isn't if ns-3 is suitable, the point is if any network simulator is suitable.

When you do a simulation, you must have a clear idea of what's the thing you want to measure/test.

Now, about your question, you can use ns-3 to do all you're asking. It's not yet in the codebase, but adding it should be easy enough. Well, it depends if you wanna do it at application level or to add the functionality in an already developed model (which could be a bit more complex).

You'll be able to simulate all the key exchange and so on. The point is: what data will you get ?

• Network overhead: yes, with great detail.
  
• Computing overhead: no, impossible to measure (because ns-3 is an event-driven simulation and it doesn't take into account (yet) the data processing delays. You could add it, but with considerable effort.
  
• Memory overhead: yes, but with a twist. How you store data (e.g., keys) is dependent on your architecture, and the data you'll have from ns-3 can be "strange". You'll have a good idea, for example, of the number of keys stored, but not about how much memory is needed (because it depends on the way you store the data).
  
• Attacking the infrastructure: not really. Attacking a key exchange or cypher is a matter of computing, not data exchange. You can successfully simulate and evaluate things like wormhole attacks, chop chop, or closing a TCP session, but evaluating how secure a crypted exchange is… hard. Im most cases it's even useless to actually crypt the data, it's enough to say "it's crypted" and have the node discard the datagram if the key isn't right.

So, the answer is: it's possible, but you have to carefully think about the data you plan to get.

Hope this helps,

T.