If you're like most people, you likely overweigh the amount of anonymity you have online. It's evident from the number of trolls we find regularly on blogs and social media, that people feel safe and untraceable when doing any kind of typing online.
That couldn't be further from the truth.
You see, people write in unique ways: everyone differs in their use of punctuation, adverbs, and complexity of words. There are thousands of nuances in everyday text.
As you can imagine, being able to detect all these elements could be disastrous for online freedom, and gives the 'identifier' a great deal to be reckoned with. Given how easy it is to capture and process huge amounts of text at respectable accuracy, it's a no-brainer that large companies have probably already mastered user identification.
How to protect yourself
Stylometry (the art of identifying the author of a piece of text), has evolved steadily. You cannot circumvent every method of identification, but it's a good idea to distort your text nonetheless.
The methods I use are as follows:
- Substitute some words in your sentence with synonyms that you seldom (never?) use. A good resource for finding synonyms is Thesaurus.com
- Randomly add and remove punctuation, for example: have a full stop on all (or none of) your lines, add more (or less) commas than normal. Remember, you are attempting to type in a significantly different way
- Intentionally use misspelled or alternately spelled words (such as color instead of colour, thirty-two instead of 32, okay instead of ok).
- Ensure your writing style on one piece of text, or on one site is consistent. Vary your writing by website instead of creating a new style for every post
- If you don't use emojis use them, if you do, don't 😛
- Be aware of cultural influence on your English and attempt to hide or at least distort it.
Thanks for reading till the end. Here are some additional resources that may help you:
Recommended PDF: Stylometry using Python
Java Graphical Authorship Attribution Program
3 approaches to stylometry with Python
[PyPi] stylometry library
'anonymized' data to practice stylometry