Energy Management in Secure Mobile Environment

1
Energy ManagementinSecure Mobile Environment

• Ramesh Karri, Piyush Mishra
• Department of Electrical and Computer Engineering
• Polytechnic University
• Brooklyn, NY

2
Outline

• Design policies for energy-efficient secure
  wireless sessions
• Develop a mobile test bed for accurate and
  efficient energymeasurements
• Benchmark power consumption and performance
  characteristics of themobile test bed
• Software implementations of DES, 3DES, AES
  encryption algorithms
• WLAN card (modes of operation RF signal power
  level)
• Software implementation of DEFLATE data
  compression algorithms
• Develop technique to minimize the energy
  consumed by a securewireless session and study
  the security vs. energy trade-offs

3
Mobile Test Bed
Spectrum24 11 Mbps network adapter card
Tektronix TDS 3054 Oscilloscope
Sycard PCI 100 extender card
?
To the Li-ion 4 V battery
Symbol PPT 2800
?
Tektronix TCP 202 current probe

• Symbol PPT 2800 SA-1110 206 MHz processor, 16
  MB RAM, 16 MB FLASHROM, 16 KB Instruction cache,
  8 KB data cache
• Tektronix TDS 3054 oscilloscope 4 channel, 500
  MHz, 5 GS/s
• Tektronix TCP 202 current probe DC to 50 MHz,
  15A DC Peak AC,Minimum sensitivity 10 mA/div,
  DC accuracy ?1 with probe calibrator

4
Secure Wireless Session

• Step 1 secure session negotiation
• client-server authentication
• session parameters exchange
• Step 2 secure data transfer
• data compression/decompression (optional)
• data authentication/verification
• data encryption/decryption
• data transmission/reception
• Step 3 session termination
• send session termination message
• store the session parameters for future use
  (optional)

5
Profiling Spectrum24 LA-4121 WLAN card
LA-4121supports five polling modes (P1-P5) and a
CAM mode
Dynamically matching the mode of operation of the
WLAN card to the application characteristics is
desirable for energy efficiency
? Stays in the CAM for this time after receiving
a frame. 2 Stays in the CAM for this time after
transmitting a frame.
6
Profiling DES, 3DES, AES on Pocket PC

• Optimized for throughput, entire code
  (instructiondata) running from cache
• 16 KB instruction cache, 8 KB write back data
  cache (32-way set associative)

1 DES, 3DES Encryption (ECB mode), 64 bit data
block, 64 bit key, AES Encryption (ECB mode),
128 bit data block, 128 bit key
7
Energy vs. Security
8
Choice of Key-exchange Algorithm
9
Protocol Optimization

• Variant 1 Server looks up client certificate
  from its own source
• Variant 2 Combine previous session association
  with newly exchangedrandom values to generate
  new session parameters
• Variant 3 Implant the secret in server and
  mobile unit
• Proposed adaptive scheme Use variant 1 for
  session negotiation and variant 2for session
  refresh

10
Software vs. Hardware Encryption
Memory access
Encryption

• Encryption in hardware
• Pros
• Faster
• Less power
• More secure (e.g. againstphysical and
  intercepting attacks)
• Cons
• Cost
• More area
• Energy efficient??

Transmission
Idle system
t1
t2
t3
t4
T
t4?
t1
t3
T
t2-?
11
S/W vs. H/W Encryption Energy
Encrypting and transmitting (P1 mode) 1000 UDP
packets (600 bytes/packet)
1 Though the actual Tx time for 600 bytes packet
is between 0.74 to 0.8 ms, system response is
sluggish by 0.25 to 0.35 ms. 2 System current
consumption during active transmission 3 DES
System WLAN card in CAM mode 4 Idle system time
deduced as Col 1 (Col 2 Col 3 Col 4)
12
Secure Session Energy vs. Throughput

• For a secure session over WLAN, energy saved by
  implementing encryption in h/w is not significant
  (16.44 for 3DES and for 3.77 for AES). This is
  because energy consumed by encryption is small as
  compared to the energy consumed by transmission
  and by the idle system
• Reception power, sleep power and transmission
  power, in that order, affect the system energy
  consumption characteristics over a long term of
  use

13
Data compression

• Compression reduces the energy consumed by data
  transaction (both transmission reception),
  cryptographic computations, idle system and
  messages exchanged during session negotiation

• Two important observations for energy efficiency
• Optimum parameter selection to tradeoff
  compression ratio vs. energy
• Matching the compression block (CB) size to the
  data cache size

CL Compression level, ML Memory level 1 Calgary
corpus benchmark
14
Data compression vs. Energy
1 Includes data authentication, data encryption
and key refreshes
15
RF Signal Power vs. WLAN card power
1 Operates at 3.3 V2 Operates at 5 V
1 Used by Spectrum LA-4131
16
Choice of RF signal power

• Power consumed by WLAN card f (RF signal
  power)
• Designing policies for dynamic RF signal power
  configuration based on WLAN card power
  characteristics, distance from the access point
  and network conditions is highly desirable