From iptables-extensions(8):
The "net.netfilter.nf_conntrack_acct" sysctl flag controls whether new
connections will be byte/packet counted. Existing connection flows will
not be gaining/losing a/the accounting structure when be sysctl flag is flipped.
This accounting information can then be used to create rules for the connbytes
netfilter module. Note that setting this option does not require a reboot, this can be enabled at runtime (for new connections only):
sudo sysctl -w net.netfilter.nf_conntrack_acct=1
When enabled, /proc/net/nf_conntrack
will carry information about packet count and size:
$ grep dport=1234 /proc/net/nf_conntrack
ipv4 [...] sport=58730 dport=1234 packets=324 bytes=480352 [...]