4 ms·
This is super cool. Very similar to timing attacks. I wonder if there is a way to tune the model. Essentially cater the probability of each key to a person with
by syntaxing 8y ago
This is super cool. Very similar to timing attacks. I wonder if there is a way to tune the model. Essentially cater the probability of each key to a person with maybe a sentence or something.
- sdenton4 8y agoYeah, this is exactly the sort of thing where augmenting the base model with a language model will have some good returns...
- ggerganov 8y agoIn keytap2 I'm trying to make use of the statistical distribution of n-grams in the language. The idea is to first group the unknown keys into clusters based on how similar they sound. The prediction then is performed by breaking the obtained substitution cypher (assuming each cluster corresponds to a letter).