Professional Documents
Culture Documents
Intelligent Transport Layer Protocols For CRN
Intelligent Transport Layer Protocols For CRN
Group Members:
Syed Waqas Ali
Maliha Rafi
Maria Khalid
Contents
Motivation
Spectrum Licensing
Introduction of CRN
Problems at different layers
Transport Layer issues
Different TCP Flavors
Sensing Mode Problem
Cognitive Radio Ad Hoc Networks
Security issues in cognitive networks
Conclusion
Motivation
Wireless networks works on fixed spectrum
allocation
According to FCC, only 15% of allocated spectrum is
being used efficiently
New paradigm needed that provides the capability to
share or use the spectrum in an opportunistic manner
Spectrum Licensing
Primary Licensing
License holder of the band operates as the “primary” user
Secondary Licensing
“Secondary” users are allowed to access the unused licensed
band of primary, complying with certain requirements
Cognitive Radio Networks
Cognitive radio is a wireless communication system which
is aware of the environment and its changes and can adapt
its transmission parameters accordingly.
Architecture of CRN
CogNets Problems at different layers
Physical Layer
Data link Layer
Networks Layer
Transport Layer
Packet loss in cognitive Radio Networks reasons:
Congestion
Bit Error
Hands Off
Sensing mode and its duration
TCP assumes Packet loss reason :
Congestion
Split-Mode (In direct TCP)
•Split wireless medium from wired medium
•Intermediate routers
•Intermediate router do the necessary
retransmissions
What if wireless device is the sender?
SPLIT: uses TCP Reno over wireless link
FH BS MH
Internet
Problem:
IPv6 encryption
Different TCP Flavors
TCP Reno
TCP New RENO
TCP SACK
TCP Vegas
TCP Reno
Fast Retransmission
Fast recovery
Problem:
The reason of packet loss
may not be detected.
Difficult to implement.
TCP
Vegas
Proactive approach
Detect congestion by
comparing sending rate
with expected rate of RTT
Handle multiple packet
losses
Problem:
Slow start increase
window exponentially
after every other RTT
Not compatible with new
Reno
Efficient TCP Flavor for CRN
TCP NEW RENO with SACK
Why not RENO?
Why not Vegas?
Sensing Mode and its duration
RTT keep increasing
since TCP is unaware of
any sensing activity
Ultimately calculated
RTO increases
Problems:
Time it takes to pass information to other layers
Spectrum switching time
Treatment of sliding window after having lower layer
knowledge is undefined
Cognitive Radio Ad-hoc Networks
Traditional Ad Hoc network issues
Congestion
Channel induced packet drops
Mobility based packet losses
Cognitive Radio Ad-hoc Networks
Cognitive Radio Ad-hoc Network issues
If an intermediate node is engaged in spectrum sensing
and is unable to forward packets
Sudden appearance of primary user can force the
secondary user to terminate its transmission and hence
increase in RTT
Spectrum Sensing State
Flow Control
T = latency/2 ---------------------------------------------------------(2)
T = time for transmitting a packet on the link (z – 2, z -1)
Latency = bidirectional latency of link
The buffer space available in the node z – 1 can be decremented at intervals T when
node z is involved in sensing. This is how the packet loss is reduced until the buffer
space is filled.
Spectrum Switching State
+1)
z+1
, z
(z
C2
C1
(z
-1,
z)
z-1
Bandwidth b = S1 + S2/ L1 + L2 ------- (1)
z-1 z
S1 = size of probe message
e l lis t S2 = size of ACK message
n
Chan L1 = link transmission delay(z, z - 1)
Chann L2 = link transmission delay(z - 1, z)
el sele
cted(c
3 )
Time RTT=RTT1 + lt1 – lt2 --------------------(2)
Old Channel
RTT1 = RTT of newly identified channel
e
Prob Lt1 = Link latency during the last normal
state of protocol
ACK Lt2 = total bidirectional link latency carried
by ACKs forwarded over intermediate hops
New Channel
cwnd = 0.8 x b x RTT ---------------------(3)
Security issues in Transport Layer
of Cognitive Radio Networks
Jelly Fish Attack
Dropping
Misordering
Delay Variance
Key Duplication
Sessions occur for a short duration
SSL and TLS establish cryptographic keys
Conclusion
Due to bursty data loss TCP new Reno with Sack
variant is better approach for CogNets
But due to their sensing activity and need to make
then cross layered, certain modifications are needed
Still an unexplored area
Thank You