Transmission Security via Fast Time-Frequency Hopping

1
Transmission Security via Fast Time-Frequency
Hopping

• PI Eli Yablanovich
• Co-PIs
• Rick Wesel
• Ingrid Verbauwhede
• Ming Wu
• Bahram Jalali

UCLA Electrical Engineering Department
2
Four users, each with four bits

• Alices Data A1, A2, A3, A4
• Bobs Data B1, B2, B3, B4
• Carols Data C1, C2, C3, C4
• Daves Data D1, D2, D3, D4

3
Random Hopping on a Time-Wavelength Grid

• A user appears on zero, one, or more
  wavelengths each symbol.
• Users select positions in grid in an
  unpredictable fashion.

A1
D2
C1
D4
Wavelength 1
A2
C2
C3
B1
Wavelength 2
D1
A4
D3
B2
Wavelength 3
C4
A3
B3
B4
Wavelength 4
Time
4
Grid-to-Grid Mapping is a Switch
User
Wavelength
Bit Index
Time

• There are 16! possible configurations of this
  switch.
• The switch configuration may be specified by
  log2(16!)44.25 bits

5
Grid-to-Grid Mapping is a Switch
16 Users (A-P)
Wavelength
Switch also supports 16 users on 16 wavelengths
with wavelength-only hopping at a total rate of
10 Gbps.
6
A Pipelined Switch

• There are 16! possible configurations (44.25
  bits).
• There are 56 switches, but four can be fixed so
  that 52 bits specify the configuration.
• Thinking about future feasibility, for a
  100?100 switch, not all switch positions need to
  be randomized.

7
Four Switches Taking Turns
155MHz
2.5Gbps
2.5Gbps
16X16 Switch
116
161
User 1
Modulator
Each 16X16 switch (the blue box) runs at 155
MHz, which is ¼ times 1/16 times 10 GHz.
l1
116
161
16X16 Switch
User 2
Modulator
l2
41
116
161
16X16 Switch
User 3
Modulator
l3
16X16 Switch
116
161
User 4
Modulator
l4
Serializer
116
161
de-Serializer
8
The Big Picture
User
Wavelength
Bit Index
Time
We need 52 bits or 9 Gbits/sec (We can do about
20 Gbits/sec)
Advanced Encryption Standard Random bit
generator (initially just a linear feedback shift
register)
9
What Kinds of Security Are Possible?

• Security by Obscurity
• This is no security at all. Obscurity is
  fleeting.
• Security by computational difficulty
• Standardized systems like DES and AES rely on
  this.
• Must consider attacks where plain-text is known.
• The one-time pad that nobody else knows
• Perfect as long as the pad remains secret.

10
Hopping versus Spreading

• Our technique focuses on the addition of
  cryptographic security in the context of
  relatively straightforward frequency-hopped CDMA.
• Certainly, similar techniques could be applied to
  the other OCDMA techniques described during this
  meeting.
• However, in every case, the real security comes
  from (high speed) cryptographic security rather
  than obscure optical techniques.

11
Network Security

• Most sophisticated security techniques add
  security at the source only (application layer).
• Our technique adds security at the physical layer
  (or at the network layer).

12
Why Have Network Security?

• Increase the difficulty of attack, even with
  plaintext available. (The ciphertext of an
  individual stream is now difficult to receive.)
• Adds security with minimal latency (the latency
  inherent in the timespan of the permutation)
  because AES processing is not in the real-time
  path..

13
Synchronous vs. Asynchronous

• Our original vision was for a system with 100
  spectral efficiency (assuming dense wavelength
  packing), but with synchronous operation (and a
  universally known key) as a requirement.
• However, our system concept can easily trade
  spectral efficiency to operate asynchronously.
  In this case each transmitter can have its own
  key. When overhead is low, collisions are rare,
  and may be handled by a light error correction
  code.
• In one scenario 5 spectral efficiency yields a
  1 bit error rate.