Just a note, I actually play a decker (Spikes) that rolls 20 dice on matrix perception. So 27 is not all that invisible at itself. The problem is device limit though as she cannot pull off so much DP to spot that decker.
20 vs 27? Even before limits, your decker only has a 19% chance to score net hits. If Spikes' rolling that on every IP she gets, the hidden decker could expect about 5 of your deckers IPs to go by before getting spotted.
I just got back home where I can put limits in the calculation. You can replicate this with the "lowest of" function in anydice to emulate limits. For example a diepool of 20 with limit 9 would be "output [lowest of [count {5,6} in 20d6] and 9]]".
Limit 4 = 0.78%
Limit 5 = 2.41%
Limit 6 = 5.54%
Limit 7 = 9.78%
Limit 8 = 13.91%
Limit 9 = 16.82%
Limit 10 = 18.30%
No limit = 19.02% chance of net hits
As you can see, Spikes' chances can get rather dismal unless she puts a pretty high stat into data processing. Using hot-sim (diepool 22), only improves her chances significantly at limits 7 and higher.