Cryptography Reference
In-Depth Information
Finally, using the QMTrace step, we can produce the subfield subcode
202111021110210200
220111102011021020
022111210101102002
202022202011210110
220202220101021011
022220022110102101
221212002102111010
122221200210111001
212122020021111100
001101202000011001
100110220000101100
010011022000110010
2 10 2 11100000000000
0
21
1 21010000000000
1
02
1
12001000000000
0
20000100000000
0 00 0 12000010000000
0 00 2 01000001000000
0 22 1 22000000100000
2 02 2 12000000010000
2 20 2 21000000001000
1 22 1 02000000000100
2 12 2 10000000000010
2
00
1
H =
,H sys =
.
21
0
21000000000001
The systematic form H sys can be computed by inverting the matrix consisting of
the last trm columns. From the systematic parity-check matrix, QMSignature
extracts those entries marked in boldface, which are sucient to describe the
public generator matrix
100000 201000022122
010000120000202212
001000012000220221
000100 211102122120
000010121210212012
000001112021221201
G sys =
.
Note that for the parity-check matrix, the signature of each monoidic block is its
first column, but for the generator, which contains transposed monoidic blocks,
the signature is the first row.
B
Decoding Square-Free Goppa Codes
For codes with degree t and its average distance at least (4 /p ) t + 1, the proposed
decoder can uniquely correct (2 /p ) t errors, with high probability. The correction
capability is higher if the distribution of error magnitudes is not uniform, ap-
proaching or reaching t errors when any particular error value occurs much more
often than others or exclusively. The parity-check matrix used by this algorithm
is in the form (1).
At some point of this algorithm, we will call the WeakPopovForm algorithm
(also present in [BLM10] and described below) to find the short vectors in the
lattice spanned by the rows of
g
00 ... 0
v 1
10 ... 0
v 2
01 ... 0
A =
,
(5)
.
.
.
.
. . .
v p− 1 00 ... 1
Search WWH ::




Custom Search