Yes, you can use such resources for linguistic aspects of the task.
The main restriction is on using information relating to the rumour from after the time of the original post, as the idea is to design systems that verify rumours as they appear.
Of course, work is more compelling where resources are shared, but a private dataset isn't against the rules.