In our example, Eve tried decoding various messages until she saw something that looked like English. Try automating this.

Get a data structure containing a few thousand English words.1

Create a program that takes in an encoded string, then try decoding it with all 25 shift values.

Use the dictionary to try to automatically determine which shift is most likely.

Because you have to deal with messages with no spaces, you can simply keep a count of how many dictionary words show up in the decoded output. Occasionally, one or two words might appear by accident, but the correct decoding should have significantly more hits.