1. Lecture 10. P2P VoIP D. Moltchanov, TUT, Fall 2014
D. Moltchanov, TUT, Spring 2008

2. Outline What is VoIP 2G VoIP system: SIP-based
Architecture
All-IP calls
PSTN-IP-PSTN call
Messages and formats
3G VoIP system: P2P Skype
Capabilities
Login to the network
Calling
Security features
Future VoIP systems: P2P SIP?

3. VoIP basics

4. What is VoIP VoIP definition
Specific sets of protocols to carry voice over the IP
IP can be private or public Internet
Early provided by operators, now by third parties
The way VoIP is implemented evolved over time
Starting from late 90s
End devices
Hardware phones
Software phones

5. Benefits of using VoIP Operational costs Flexibility
No need for two separate networks (voice and data)
Rich set of free features e.g. conference, forwarding…
Lower costs since only Internet access is needed
Compare pricing scheme: flat for IP vs. per minute
Flexibility
More than one call over a connection
Calls can be encrypted
Location independent
Integration with other services
Computer + phone
Note: Internet to Internet calls are free…

6. VoIP evolution

7. 2G VoIP systems: SIP

8. 2G systems: protocols Data transmission: data plane
Real-time transport protocol (RTP)
Real-time control protocol (RTCP)
Signaling: control plane
H.323
Session initiation protocol (SIP)
Location service: control plane
Proxies
Finding via DNS SVR
PSTN-IP gateways: control plane
Media gateway control protocol (MGCP)
H.248 (similar to MGCP)
Description of codecs used: control plane
Session description protocol (SDP)

9. 2G VoIP infrastructure: SIP+RTP
SIP+RTP makes a perfect combination
DNS is for proxy resolution

10. SIP What it does Properties Functionality
Setup, control, terminate calls
End-to-end signaling
Allows for end system and MG control signaling
Properties
Partially distributed
Text based (you may read messages in e.g. Wireshark)
Simple (simpler than H.323)
Limited but sufficient functionality (i.e. optimized)
Functionality
Point-to-point and multipoint calls
Additional features
URL are used for addresses (e.g.

11. Basic call: all-IP

12. Basic call: PSTN-IP-PSTN

13. Reality is a bit more complicated
How do we know where to call to?
All parties need to be registered with SIP servers
Which server to register with?
Multicast to all known SIP proxies “sip.mcast.net” ( )
How to find the person?
SVR (service record) in DNS servers for proxy resolution
Specifies the location of proxy server for a specific service
That’s why we had DNS servers there
What is about type of the media?
I use G.711, G.723, G.729, your client may prefer G.729
Use SDP (session description protocol)

14. SIP calling

15. Requesting presence info
Alice wants to be informed when Bob is online

16. Registering and notifying
Bob goes online, Alice is notified

17. SIP messages: RFC 3261 REGISTER INVITE ACK CANCEL BYE OPTIONS
Register and notify proxy about its IP and URLs
INVITE
Establish a media session between users
ACK
Confirms reliable message exchanges
CANCEL
Terminates a pending request
BYE
Tears down a session between two users in a conference
OPTIONS
Solicits info about the capabilities
SUBSCRIBE
Get notifications

18. SDP: RFC 2327 Media streams Addresses Ports Payload types
A session can include multiple streams of differing content.
SDP currently: audio, video, data, control, and application
Addresses
Indicates the destination addresses
Ports
UDP ports for each stream
Payload types
Media format for each stream
Start and stop times
For broadcasting (radio, video)
Originator
For broadcasting

19. 2G systems: what we see? SIP almost makes a P2P system. Why?
Data connection goes directly between parties
Control is still performed using servers
Similar to “almost P2P” platforms?
Music sharing (actually, file sharing) systems of late 90s
Napster
Audiogalaxy
Audiogalaxy example
You login using a specific agent
Agent imports your song titles to the common DB
Once you need something you search through this DB
Result is IP of the systems holding this title

20. 3G VoIP systems: P2P Skype

21. What is Skype? A P2P application for Proprietary signaling
Primarily, VoIP
Video calls added recently
Instant messaging
File transfers
Proprietary signaling
Negotiable voice codecs
Proprietary encoders
Gateways to/from PSTN/SIP
Interesting features
Works well in almost all network conditions
Works across NATs and firewalls

22. Overlay network Consists of Client Skype clients (SCs)
Supernodes (SNs)
Login server
HTTP server
Client
Used to make activities
Stores configuration
Connected to some SNs
Stores few SN addresses
Updates them periodically

23. Functionality of elements
Supernode
A node that can accept incoming TCP connections
Preferably has enough CPU, memory, and BW
Preferably not behind firewall or NAT
There are default supernodes
Do signaling directly with other SNs
Sometimes perform signaling and data transfer for SCs
Login server
Ensures that names are unique
Authentication point
HTTP server
Used for updates

24. Skype features Codecs Ports Coding and encryption Host cache
Default: wideband 16KHz sampling, 5Kb per direction
140pcks/s., 67 bytes of payload
Ports
80 HTTP, 443 HTTPS TCP
Random UDP ports
Coding and encryption
Everything is encrypted using AES
Keys: 256 bits, symmetric
RSA for exchanging the symmetric keys
Host cache
Supernode list (IP, port) to fasten up the process, 200 entries
Updated periodically, some SNs are always there

25. Login: joining the overlay
Contacting central servers
Authentification/authorization
Looking for updates at HTTP
Joining the overlay
Refresh of SN list (shared.xml)
List containing SNs
Sends UDP packet to a default SN
Chooses a certain SN
Opens TCP with connection with this SN
Connection is maintained throughout a session
Exchanges info on on-line nodes
Testing for SN capabilities
Client sometimes checks whether it could be a new SN

26. Login: firewall blocking
If firewall blocks UDP for SN list refreshing
Establishes TCP connections with few SNs
Gets info on the SN list
All but one connections are torn down
If firewall blocks connection to the login server
Uses SN as a relay to authenticate

27. Calling Signaling Procedure is as follows
Using TCP connection directly
Overlay if impossible otherwise
Media is carried using UDP
Procedure is as follows
A queries SNs for the address of B
Once obtained signaling is done directly using TCP
Then voice is carried using UDP

28. Calling: firewall blocks UDP
Signaling by SNs on behalf of users
Media: via TCP using 4 SNs as relays

29. Calling: port-restricted NAT
Step 1
User A gets address of SN of B
Sends UDP query containing its external address
SN of B replies with external address of B
Step 2
A and B establish UDP flow using hole punching
A and B establish TCP connection using 4 SNs as relay

30. Calling: symmetric NAT
Step 1
A obtains the address of B
Step 2
Trying hole punching
It does not work (symmetric NAT)
Step 3
They use TCP via 4 SNs as relays

31. Some facts Skype completely fails when firewall blocks UDP and TCP
This is exceptionally rare
Mechanism 1: hole punching
Mechanism 2: tunneling via SN
Mechanism 3: use of overlay
Trying to use direct connections as much as possible
SNs may tear down
Nodes may overload overlay
Skype is seen as unwanted by enterprises
May traverse firewalls, NATs
May bring unwanted stuff

32. Security: self-security
It is not open source!
Binary code
Parts of the binary are encrypted
Code is decrypted in memory in run-time
Contains checks for
Presence of a debugger
Code modification
Result: stops of crushes

33. Security: network security
Facts
Uses a proprietary protocol
Encrypts all the traffic
Signaling packets
Payload: encrypted using RC4 stream
RC4 key can be recovered from the packet
VoIP packets
Encrypted using AES
Only sender and receiver can decrypt

34. Future: P2P SIP?

35. What do we want? A bit more competition + open architecture
Goals and motivation
Get rid of SIP servers
No fixed topology
Audio/video/IM all-in-one
Interoperability with SIP
Possible
Join DHT
Query position in DHT
Update neighbors
Repeat periodically
More info at: