1. Neha Jain Shashwat Yadav
VoIP Encryption
Neha Jain
Shashwat Yadav
ECE Advanced Internetworking

2. Context for the Problem
VoIP  Voice over Internet Protocol
Real time transmission of voice signals as packetized data over an IP based network
Results from preliminary work indicated
Encryption incurs delays
Delay not significant enough to affect voice quality
What causes encryption delays to worsen.
ECE Advanced Internetworking

3. ECE 4605 Advanced Internetworking
Problem Statement
To measure one way delays in a VoIP session, incurred before and after Encryption.
Evaluating the effect on VoIP quality under varying conditions which cause increasing delays
Increasing bandwidth contention
Increasing CPU usage
Increasing the application transmission rate
ECE Advanced Internetworking

4. ECE 4605 Advanced Internetworking
Delay Budget
Packetization Delay
Overhead vs. Delay
Codec Delay
Bandwidth vs. Quality
Jitter buffer delay
Collects packets and passes it in a regulated, sequenced order
Propagation, Transmission, Queuing delays }
Delay Budget
0 – 150 ms: Acceptable
150 – 250 ms: Perceptible, but OK
250 – 400 ms: Low quality
>400 ms: Unacceptable
Degradation in voice quality if encryption delay exceeds delay budget
ECE Advanced Internetworking

5. Crypto-engine Bottleneck
Scenario - I
Scenario - II
App. Layer
Generates every
50 ms
App. Layer
Generates every
50 ms
Network Layer (IPsec)
Network Layer (IPsec)
Requires 20ms for
encryption algorithm
Requires 70ms for
encryption algorithm
20 ms additional fixed delay
No longer fixed to 20 ms
Crypto-engine bottleneck depends upon transmission rate of application layer and on time required for encryption.
ECE Advanced Internetworking

6. ECE 4605 Advanced Internetworking
Work Outline
Factors affecting voice quality
Delay Budget
Crypto-Engine Bottleneck
1) One way delay
measurement using
Ethereal time-stamp
2) Presence
of contending
flows in the
network
3) Increasing
C.P.U
utilization
4) Different application
transmission rate
to network layer
ECE Advanced Internetworking

7. Work done :One Way Delay
Used Ethereal captures to find absolute times 
Used IP sequence numbers to synchronize the two flows 
Use of absolute time scales
- No synchronizing of clocks required
- But it provides only the increase in delay due to encryption.
ECE Advanced Internetworking

8. ECE 4605 Advanced Internetworking
One Way Delay : Result
Delay induced after encryption
DES = 3.67 ms
3DES = 4.54 ms
ECE Advanced Internetworking

9. Contending Flows : Result
Delay induced after encryption
DES= 4.58 ms
3DES = 7.92 ms
But, there was a high perceivable delay in voice transmission
Change in absolute delay
3DES= ms
DES= ms
No= ms
ECE Advanced Internetworking

10. ECE 4605 Advanced Internetworking
Delay Analysis
Application processing delay
Application Layer
(VoIP)
Application Layer
(VoIP)
Ideal Measurement
Processing Power
Transport Layer
Transport Layer
Network Layer (IPSec)
Network Layer (IPSec)
Encryption/ Decryption delay
MAC Layer (Ethereal)
MAC Layer (Ethereal)
Measured Delay
Explains the discrepancy between measured delay and perceived quality. We only measure network delays
ECE Advanced Internetworking

11. ECE 4605 Advanced Internetworking
Increasing CPU usage
CPU usage was kept constant at 100% by running multiple applications
First – VoIP call monitored, observed packet loss and significant delays. No Jitter.
Used ITG (Internet Traffic Generator)
Client – Server traffic generator application for Windows!!!
Allows RTT calculation, different bit rates
DEMO
Second – TCP flow was generated and RTT measured without VoIP (avoid UDP interference).
ECE Advanced Internetworking

12. ECE 4605 Advanced Internetworking
Increasing CPU usage
Significant degrade in Voice quality for all 3 cases.
DES: ms
3DES: ms
Increase can be attributed to encryption delay, increased by reduced CPU processing power.
Still doesn’t take into account delay due to application layer processing.
ECE Advanced Internetworking

13. Application transmission rate
Second test to look into crypto-engine bottleneck
As transmission rate increases, it creates a bottleneck at network layer
Generated TCP flows at varying rates.
Measured Rtt.
Rtt increased as sending rate increased.
Anomaly in lower sending rates.
ECE Advanced Internetworking

14. ECE 4605 Advanced Internetworking
Implication
Delay incurred by turning on encryption doesn’t seem to have an effect.
Decreasing the amount of CPU processing power definitely affects voice quality. So when using VoIP phones of a limited capability, this is a problem.
IPSec encrypts all traffic, making it a serious bottleneck in event of limited CPU processing power. In future, encrypting only the VoIP application would cut down the delay.
ECE Advanced Internetworking

15. Spam Over Internet Telephony (SPIT)
Encryption prevents eavesdropping.
However, no safeguarding against spam.
In one of our test runs, we faced one such issue.
ECE Advanced Internetworking

16. Problems Faced/Future Work
Lack of a software that would only encrypt a particular application with varying key size.
No way of measuring one-way delays, due to unsynchronized clocks of the two computers.
Windows and Linux incompatibility.
Future work:
Measuring application to application delays.
Using gradually increasing CPU processing rates.
Using higher encryption key sizes and different algorithms.
ECE Advanced Internetworking