The Python AI module in Wesnoth 1.4.x and 1.5 before 1.5.11 allows remote attackers to escape the sandbox and execute arbitrary code by using a whitelisted module that imports an unsafe module, then using a hierarchical module name to access the unsafe module through the whitelisted module.

More... (http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2009-0367)