I can't access the research paper, but it seems one of their findings is that miners speculate on the market so fee is not their biggest incentive to mine to some extent, which is why calculating the cost to attack is a bit difficult. They also conclude that 51% is not feasible or too expensive if you count the possibility of the network fighting back with a fork etc. I'm not sure how reliable this is, though, since I can't verify it and the news link basically paraphrases their abstract with no in-depth analysis.