Piggybacking on comments regarding the authenticity of Pliny's letter 10.96 to Trajan, here is a little more background to use of n-grams as compared to other means to identify authorship:
Efstathios Stamatatos, Author Identification Using Imbalanced and Limited Training Texts (Proc of the 4th Intl Wkshp on Text-based Info Retrvl, 2007)
http://www.uni-weimar.de/medien/webis/e ... g-sets.pdf
Efstathios Stamatatos, A Survey of Modern Authorship Attribution Methods (JAIST 60-3, 558-556, 2009)
http://citeseerx.ist.psu.edu/viewdoc/do ... 1&type=pdf
Anđelka Zečević, N-gram Based Text Classification According To Authorship (Student Research Workshop, pp. 145–149, 2011)
It seems that it works best with identifying the author of programming source code, and identifying the authors of e-mail, but like all the methods in use today, must be "trained" and here is where the weakness lies. We don't have a corpus of works we can be 100% confident were written 100% by Pliny, of Seneca, or Cicero, etc., but we can do that with code written by programmers (this means you, wonk!).
The web nowadays hates the concept of associating dates with texts and other files preserved online, but I researched when the articles above were written and where, and included some publication data
above. Also, it doesn't help that many of the articles on Pliny's letter 10.96 compared to his other letters in Book 10 are in Pay-firewall environments. Even scribd.com seems to be running scripts as bad as anything on a site like Huffpost, the Sun, etc. Maybe it is my 9 yr old old computer, I don't know. Any smartass who suggests that I join the 21st century and get a newer computer, will find my electronic hand reaching through the internet connection to slap you silly.
Edit 1/14/18: Researched when the articles above were written and where, and included some publication data