EDA　tools　almost　completely　automate　the　current　chip　design.　However,　EDA　tools　are　all　provided　by　third-party　companies,　therefore　the　credibility　of　EDA　tools　and　the　trust　issues　brought　by　their　process　libraries　have　attracted　people＇s　attention.　Given　these　problems,　this　paper　proposes　a　self-training　hardware　Trojan　confrontation　framework　based　on　machine　learning　embedded　in　EDA　tools.　The　framework　focuses　on　the　gate-level　netlist　files　generated　during　the　integrated　chip　comprehensive　optimization　phase,　thereby　detects,　locates　and　deletes　hardware　Trojans　that　may　appear.　The　experimental　results　show　that　the　framework　can　achieve　more　than　85%　detection　effect　for　unknown　hardware　Trojans.　Moreover,　for　known　hardware　Trojans,　this　framework　can　achieve　almost　100%　detection.　Accordingly,　hardware　Trojans　can　be　located　and　deleted.　©　2019　IEEE.