"... In PAGE 7: ...6%). The per- centages of match, desirable and undesirable change are shown in Table4 (set 2, row 1). They are compa- rable with the expert BN 0-N and only slightly worse than the expert BN H/M/L results.... In PAGE 7: ... In this case LZEs were all grouped with ATEs, as were AMOs. The match results are shown in Table4 (set 2, rows 2 and 3). Clearly, summarising the results of 24 DCT into types gives relatively poor performance; it is proposed that this is because many pairs of the classes are dis- tinguished by student behaviour on just one item type, and SNOB might consider these di erences to be noise within one class.... In PAGE 7: ... The data was randomly divided into ve 80%-20% splits for training and testing; the training data was used to parame- terise the expert BN structures using the Netica BN software apos;s parameter learning feature5, while the test data was given to the resultant BN for classi cation. The match results (averaged over the 5 splits) for the ne classi cation comparison of the expert BN struc- tures (with the di erent type values, 0-N and H/M/L) with learned parameters are shown in Table4 (set 3), with corresponding prediction results (also averaged over the 5 splits) given shown in Table 5 (set 2). The average prediction probabilities for the BN with learned parameters are better than for the expert BNs for the O-N type values (0.... In PAGE 8: ....4 to 2.2, while the number of parameters varies from about 700 to 144,000; the structures produced for the H/M/L data seem simpler using these measures, but this is not statistically signi cant. The percentage match results comparing the CaMML BN classi cations (constrained and unconstrained, O- N and H/M/L) are also shown in Table4 (sets 4 and 5), with the prediction results shown in Table 5 (sets 3 and 4). The prediction results for both 0-N and H/M/L are similar to those of the fully elicited expert BNs.... ..."

"... In PAGE 8: ...atrix in DES, s3DES and s5DES is found to have greater value than 0.6. Therefore we can infer that Boolean functions consisting of DES, s3DES and s5DES do not satisfy the SAC. Finally we checked the uniformity of linear distribution table of a S-box as shown in Table5 . We 2The name of s4DES is skipped with intention since it was distributed in an informal way.... ..."

"... In PAGE 5: ... When the US export laws were in e ect, the export key length was arti cially reduced to 40 bits, making DES even less secure. Table7 shows the DES usage in detail. While the old US export regulations no longer apply, almost 67 percent of the servers surveyed still support these weak keys, and most Cipher Number Percentage DES-40 12930 66.... ..."

"... In PAGE 6: ... In this Section, we compare the quantitative characteristics of S-boxes in DES and siDES in various points of cryptographic view and evaluate the goodness-of- t of them. We checked the nonlinearity of 4 Boolean functions: Z6 2 ! Z2 consisting of an S-box as shown in Table1 . In the output bit column of this table, 4 denotes the most signi cant location of an output vector and 1 denotes the least signi cant location of an output vector.... ..."

"... In PAGE 21: ....2.2.7 Overview of Measurement for Every C-module This section presents numbers for processing times (in ms) for all algorithms implemented in separate C-modules. Table 5 depicts MD5, Table 6 includes MD4, Table7 shows DES, Table 8 presents IDEA, and Table 9 covers RC5 with 12 rounds and 128 bit keys. A graphical Table 5.... ..."

"... In PAGE 21: ....2.2.7 Overview of Measurement for Every C-module This section presents numbers for processing times (in ms) for all algorithms implemented in separate C-modules. Table 5 depicts MD5, Table 6 includes MD4, Table7 shows DES, Table 8 presents IDEA, and Table 9 covers RC5 with 12 rounds and 128 bit keys. A graphical Table 5.... ..."

"... In PAGE 4: ... n283n29Receiving RM cells n28lines 22n7b29n29: for a forward RM cell, record its contents and forward it to the downstream node; for a backward RM cell, update the local credit-balance by CU contained in the RM cell, n0cll in the RM cell with local count and CI bit, and then send it to the upstream node. The Destination Node Algorithm n28 Table3 n29. Two 00.... ..."