A number mathematically calculated for a packet by its source computer, and then recalculated by the destination computer. If the original and recalculated versions at the destination computer differ, the packet is corrupt and needs to be resent or ignored.
The mathematical procedure for performing a CRC is specified by the International Telecommunication Union (ITU) and involves applying a 16-bit polynomial to the data being transmitted by the packet for packets of 4 KB of data or less, or a 32-bit polynomial for packets larger than 4 KB.
The results of this calculation are appended to the packet as a trailer. The receiving station applies the same polynomial to the data and compares the results to the trailer appended to the packet. Implementations of Ethernet use 32-bit polynomials to calculate their CRC.