为了降低TCP流的处理开销,可以从TCP/IP报文中选取某些位串来作为流的标识.从位熵和随机性的角度分析了TCP流首报文的顺序号(SN)和确认号(AN)的分布,提出了一种从重尾的IP或TCP流里获得随机均匀的流标识的新方法(FIDSAN).实验结果表明,在可以接受的冲突概率下,TCP流首报文的顺序号和确认号的部分高位比特可以用来作为流标签.给定冲突概率时,该流标识的比特长度可以根据一个由观察窗口和流ID值域导出的关系式求出.与TCP五元组,CRC,Checksum等比较发现,FIDSAN具有更低的计算开销. In order to reduce the processing overhead of TCP stream, some bit strings can be selected from the TCP / IP packets as the stream identifiers. From the perspective of bit entropy and randomness, the sequence number (SN) and acknowledgment (AN), a new method to obtain stochastic uniform flow identification from heavy-tailed IP or TCP flows (FIDSAN) is proposed.The experimental results show that under the condition of acceptable collision probability, The sequence number and the part of the high order bit of the acknowledgment number can be used as the flow label. Given the collision probability, the bit length of the flow identifier can be found from a relation derived from the observation window and the flow ID field. Group, CRC, Checksum and other found that FIDSAN has a lower computational overhead.