Perhaps you should consider using Materialized View Replication.
Instead of writing triggers, you put a materialized view log on your source table. Then you can turn your backup table into a materialized view that is based on the source table. When you perform a fast refresh of the backup table MV, it will read the MV log on the source table. Refreshes should take almost no time at all to process.
If the data must be absolutely up-to-date at all times, you can use refresh on commit so that the MV is always in synch with the table. Normally I recommend against using refresh on commit, and instead recommend a short refresh interval.
You can use triggers as you suggest, but it seems simpler to use something like MVs to create a shadow table.