Oppressive governments are increasingly leveraging the affordances of social media to harass the online opposition beyond their borders. High-profile dissidents in exile are often on the receiving end of states' online attacks because of their influential role in transnational advocacy networks. With connections at home and abroad, they are in a unique position to relay information between the domestic opposition and international audiences. To better understand how states respond to exiled activists within the online opposition, this paper outlines and evaluates a method for detecting targeted harassment in foreign influence operations. I propose using transfer learning with labeled Twitter data on offensive language detection and state-sponsored disinformation to identify tweets from targeted online harassment campaigns by foreign state actors. I test and evaluate this approach with data from Twitter's transparency reporting on information operations and tweets that mention high-profile exiled dissidents from China and Iran. I demonstrate the utility of this method by examining how state-sponsored online harassment campaigns differ from normal Twitter activity across political contexts. This work demonstrates the effectiveness of transferring knowledge between related text-as-data approaches for the task of detecting targeted harassment in foreign influence operations, which will enable further investigation into how online harassment campaigns factor into state strategies for transnational repression.