6.1 Experimental Result

Computational Overhead
In the proposed scheme, the most computationally expensive operation is the point multiplication. In order to reduce the computation cost, we pre-compute an enumerated list of multiplication on point α and pre-load this list in the smart tag at manufacturing time. When the RFID reader initiates a request, the smart tags picks a point in the list corresponding to the parameters s_u and λ and responses back to the reader. In our experiment, With the pre-computation technique, the we have measured authentication spends 140 milliseconds in average out of 100 tries.
Storage Consumption
Like we described in section 5.2, we preloaded some lists of points for each tag at the time, when the tag is made. We compare the storage space needed, when the size of the lists increases. In table 1, we show the of RAM size required increases with the respect to the growth of the list for points < α_u,β_u,0,β_u,1 >.

The next table 2 below measures the code size of gateway sensor and smart tag sensor required:

Communication Overhead
Under WM-ECC specification, each element relies on the size of finite field on the elliptic curve, which occupies 160 bits (20 bytes). Moreover, each elliptic curve point, consisted of two elements for x and y coordinate,s consumes 320 bits (40 bytes) of memory in total. In order to reduce the bandwidth, we apply the point compression technique for a transmission packet that involves elliptic curve points in communication. Once the compressed data is retrieved on the reader side, it is immediately decompressed back to the original form for further authentication evaluation. In our experiment, the size of payload, built by three EC points < α_u,β₀,β₁ >, is dramatically brought down to 112 bytes. This size reduction is smaller than the maximum payload size of 128 bytes, as specified by IEEE 802.15.4 standard. Therefore, the proposed scheme satisfied the constraint of low communication expense.