As highlighted by the figure, NAT does have some drawbacks. The fact that hosts on the Internet appear to communicate directly with the NAT-enabled device, rather than with the actual host inside the private network, creates a number of issues.
One disadvantage of using NAT is related to network performance, particularly for real time protocols such as VoIP. NAT increases switching delays because the translation of each IPv4 address within the packet headers takes time. The first packet is always process-switched going through the slower path. The router must look at every packet to decide whether it needs translation. The router must alter the IPv4 header, and possibly alter the TCP or UDP header. The IPv4 header checksum, along with the TCP or UDP checksum must be recalculated each time a translation is made. Remaining packets go through the fast-switched path if a cache entry exists; otherwise, they too are delayed.
Another disadvantage of using NAT is that end-to-end addressing is lost. Many Internet protocols and applications depend on end-to-end addressing from the source to the destination. Some applications do not work with NAT. For example, some security applications, such as digital signatures, fail because the source IPv4 address changes before reaching the destination. Applications that use physical addresses, instead of a qualified domain name, do not reach destinations that are translated across the NAT router. Sometimes, this problem can be avoided by implementing static NAT mappings.
End-to-end IPv4 traceability is also lost. It becomes much more difficult to trace packets that undergo numerous packet address changes over multiple NAT hops, making troubleshooting challenging.
Using NAT also complicates tunneling protocols, such as IPsec, because NAT modifies values in the headers that interfere with the integrity checks done by IPsec and other tunneling protocols.
Services that require the initiation of TCP connections from the outside network, or stateless protocols, such as those using UDP, can be disrupted. Unless the NAT router has been configured to support such protocols, incoming packets cannot reach their destination. Some protocols can accommodate one instance of NAT between participating hosts (passive mode FTP, for example), but fail when both systems are separated from the Internet by NAT.