Under manipulations, are some AI models harder to audit? | IEEE Conference Publication | IEEE Xplore