blog/content/posts/commits/2026-07-07-commit-78eeea5.md

title, date, draft, tags, categories, summary, author

title	date	draft	tags	categories	summary	author
[bojemoi] fix(ml-threat): break label/feature circular dependency, oversample pwned hosts	2026-07-07T23:55:30+02:00	false	commit bojemoi main	Git Activity	Commit 78eeea5 par grafana-watcher dans bojemoi	grafana-watcher

Commit 78eeea5

Repository	bojemoi
Branch	main
Author	grafana-watcher
Hash	78eeea58d13e88214844d9b11f2ee4d04d24d437

Description

• fetch_msf_data: query ALL pwned hosts exhaustively + benign_ratio*n_pwned benign sample — avoids 5/50000 imbalance that made accuracy=1.0 trivial
• build_feature_matrix: feature[6] = filtered_port_count (replaces malware_hits) confirmed_vulns used only for labels, not in X — breaks circular dependency
• train.py synthetic: feature[6] = filtered_port_count to match real data layout
• train_from_msf signature: benign_ratio param instead of msf_limit
• Result: 184 pwned + 1000 benign, test_acc=0.933 (was 1.0 trivially)

Co-Authored-By: Claude Sonnet 4.6 noreply@anthropic.com

Files Changed

M	ml-threat/train.py
M	ml-threat/train_from_msf.py

Diff Summary

 ml-threat/train.py          |  22 +++----
 ml-threat/train_from_msf.py | 142 +++++++++++++++++++++++++++++---------------
 2 files changed, 106 insertions(+), 58 deletions(-)