Information or influence operations have been used to manipulate crowds and often involve mass, coordinated inauthentic behaviors in social media. Although these operations can take many different forms and aim at different purposes, to succeed, different types of online accounts are used or mobilized: automated accounts, known as social bots; deceitful human-driven accounts, like trolls; or even ``willing but unwitting'' unaware humans. Here, we propose a large-scale quantitative analysis, which relies on datasets released by Twitter and researchers in recent years, to characterize different actors involved in recent information operations. In particular, we represent each account with a large number of features belonging to three different dimensions: credibility, initiative, and adaptability. Then, we apply dimensionality reduction to project users onto a low-dimensional space, followed by clustering, which allows us to find groups of similar accounts. We experiment with different combinations of 2 parameters that affect dimensionality reduction and clustering. In our best combination in terms of effectiveness, we obtain high-quality clusters, with 0.9 purity. Moreover, we analyze our results qualitatively by visualizing and studying clustering results with the dimension taken into account. Our results contribute to the understanding of the characteristics of the different types of actors in a social network like Twitter and increase our understanding of state-backed trolls which so far received limited attention from quantitative researchers.