Scrum should not be set-up as an excuse for bad code or for defect leakage. If there is leakage, that means there will be items being added back into your backlog and unhappy clients. Most teams that I talk to using scrum also try to use techniques from xtreme programming, like test-driven development, junit tests, automated regression test scripts. No matter the tools, you should have QA resrouces booked 100% on the team, who are ultimately accountable for their processes around issue management and the quality of the deliverable that is deployed at the end of each sprint.
Another thought: if defect leakage is a metric expected by the enterprise, you could consider treating it as a criteria on a user story for sprint 0. But, if it's not important to the project or enterprise in general, you probably won't get it from the team during execution, or the team might implement it when problems arise in an adhoc, tactical manner.