This　paper　presents　a　novel　image　authentication　scheme,　which　aims　to　identify　the　tampering　and　manipulations　exerted　on　the　digital　images.　By　introducing　the　intrinsic　dimension　estimation　of　manifold　learning,　the　proposed　algorithm　extracts　a　signature　composed　of　the　intrinsic　dimensions　for　pixels　of　the　digital　image.　Based　on　this　signature,　whether　the　test　images　are　manipulated　or　tampered　can　be　decided.　For　a　manipulated　image,　the　tampered　region　can　also　be　estimated.　In　this　way,　both　the　security　and　the　integrity　of　digital　images　are　guaranteed　and　protected.　Unlike　current　authentication　schemes,　the　proposed　method　uses　intrinsic　dimension　estimation　to　capture　and　locate　the　tampering　or　manipulations　on　the　pixel　level.　Therefore,　the　proposed　authentication　scheme　is　more　precise　and　effective.　Experimental　results　demonstrate　the　performance　of　the　scheme.　©　2024　IEEE.