TB003

DS91002A-page 2

Ó 1996 Microchip Technology Inc.

Once the system has responded to a valid code, about

65,000 valid codes will have to be received before the

same code will be used again. If the remote control is

used eight times daily, 22 years will pass before the

system responds to the same code again - once!

Therefore, a retransmitted code (like when a code

grabber is used) will never activate the system.

chronization techniques. The system will continue to

function even if the transmitter is activated repeatedly

while not in range of the receiver (as would happen if a

child played with the remote control). If a button is

pressed out of range more than 16 times, synchroniza-

tion will be lost. However, two successive transmis-

sions in range will restore synchronization. When no

response occurs to a transmitter operation, the user's

natural reaction is to press the button a second time.

Synchronization will be restored when he does. Opera-

tion is totally transparent — the user may not even

become aware that synchronization has been lost and

restored.

These obvious security benefits are attainable at a cost

comparable to existing fixed code systems. The revolu-

implementation, and reduced external component

requirements help to make this level of cost effective-

ness possible.

length of 32 bits and a key length of 64 bits. The algo-

rithm is characterized by a very economical hardware

implementation, while retaining a level of security com-

parable to Data Encryption Standards (DES). This level

of security makes it eminently suitable for code commu-

nication applications such as code hopping antitheft or

access control devices.

Information regarding transmitter identity and synchro-

nization is encoded so as to render it unintelligible to an

outsider. For decoding, it is necessary to have the

same 64-bit key originally used for encoding. There-

fore, even though the decoder (which has the key) can

identify the transmitter unambiguously, an outsider

(who does not have access to the key and/or the algo-

rithm) can glean no information at all from the transmis-

sions.

As it is impossible to insert information into the system

from outside, strategies used to attack FEAL and other

DES-like ciphers are not usable against this system.

ble for a potential assailant to predict the next code that

will be transmitted by a valid transmitter. Even if the

assailant makes a reasonable guess regarding the way

in which transmitted information changes with each

transmission, the algorithm obscures this information

sufficiently that the next code can not be anticipated. In

particular, even if the transmitted information (before

encoding) differs only in one bit from the information in

the previous transmission, the next transmission will be

totally different.

Checks exist that can be used to verify the security

characteristics of an encoding algorithm and, in this

instance, to determine whether the next transmitted

code is predictable to any degree. The Avalanche

Effect and a subset thereof, the Strict Avalanche Crite-

results give a good indication of the security offered by

the system.

• Avalanche Effect (AE)

A block cipher satisfies the AE if changing one bit

of the information causes, on average, half of the

algorithm, this implies that changing one bit in the

function and/or synchronization information will

cause an average of 16 of the 32 bits in the trans-

mitted code to change.

• Strict Avalanche Criterion (SAC)

The SAC requires that, if one bit of the encoded

information is changed, each bit in the output

must have a chance of 0.5 of changing as well.

Consequently, the probability of guessing any one

bit correctly is 0.5, and the probability of guessing

an entire 32-bit string correctly is one in about

4,300,000,000!

The tests involved using a random 64-bit key and

a Gray counter (starting at zero) as input to the

algorithm. In each case, the output was compared

with a reference (SAC) or with the previous code

(AE). In both cases, the results were as expected:

For the AE, the average number of bits changed

was 16.0 (50%), with a standard deviation of 2.83

(8.8%). For the SAC, each individual bit changed

an average of 50% of the time, with a standard

deviation of 8.8%.