Catcrafts/fingerprintd
Watch
2
0
Fork
You've already forked fingerprintd
 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Spend the presses on the open question, and report what learning actually folded

Browse source 
Two changes to the measurement, both about not wasting a finger.

The wrong-finger control is off by default now. Across every run this project
has made the wrong finger has matched zero times out of forty-odd taps; the
question that is still open is the false-negative rate, and each control tap
spends a press that could have measured it. Pass a count to put the control
back, and the summary says plainly when it was not run rather than printing a
zero that reads like a result.

And a trend run now reports how many frames learning folded, out of the daemon's
own transcript, alongside the container size. Without that a run cannot tell
"learning fired and did not help" from "learning never fired" -- which is exactly
the confusion that let the first trend run measure a static template three times
and look like a result.

Default trial length goes to fifteen taps, since dropping the control freed the
presses and ten was too few to separate three-in-ten from five-in-ten.
This commit is contained in:
Jorijn van der Graaf 2026-09-04 23:43:47 +02:00
commit edb0d8e530
2 changed files with 29 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

14
packaging/fptrial.sh
View file

@ -23,12 +23,20 @@ run() { # $1 = label
}
echo "labelled trial -> $OUT"
i=1; while [ $i -le $C ]; do run "CORRECT $i/$C"; i=$((i+1)); done
echo; echo "===== now switch to a DIFFERENT finger ====="; sleep 3
i=1; while [ $i -le $W ]; do run "WRONG $i/$W"; i=$((i+1)); done
# The wrong-finger half is skippable, and skipping it is the right default once
# the question it answers is settled. Across every run this project has made the
# wrong finger has matched ZERO times out of forty-odd taps; the open question
# is the FALSE NEGATIVE rate, and each wrong-finger tap spends a press that
# could have measured that instead. Pass a count to put the control back.
if [ "$W" -gt 0 ]; then
echo; echo "===== now switch to a DIFFERENT finger ====="; sleep 3
i=1; while [ $i -le $W ]; do run "WRONG $i/$W"; i=$((i+1)); done
fi
echo
echo "===== summary ====="
awk '$1=="CORRECT"{c++; if($3=="verify-match")cm++; ct+=$4}
$1=="WRONG" {w++; if($3=="verify-match")wm++; wt+=$4}
END{printf " correct finger: %d/%d matched (false negatives %d) mean %d ms\n", cm,c,c-cm,(c?ct/c:0);
printf " wrong finger: %d/%d matched (false ACCEPTS %d) mean %d ms\n", wm,w,wm,(w?wt/w:0)}' "$OUT"
if (w) printf " wrong finger: %d/%d matched (false ACCEPTS %d) mean %d ms\n", wm,w,wm,wt/w;
else printf " wrong finger: not run this trial\n"}' "$OUT"
echo " per-tap log: $OUT"