{"id":"W4323972003","doi":"10.48550/arxiv.2303.05498","title":"Mark My Words: Dangers of Watermarked Images in ImageNet","year":2023,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Digital Media Forensic Detection","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Banting and Best Diabetes Centre, University of Toronto; European Commission; Ministry of Science and ICT, South Korea; Berlin Center for Machine Learning; Deutsche Forschungsgemeinschaft; Institute for Information and Communications Technology Promotion; Korea University","keywords":"Computer science; Artificial intelligence; Watermark; Feature (linguistics); Variety (cybernetics); Class (philosophy); Spurious relationship; Extractor; Digital watermarking; Pattern recognition (psychology); Image (mathematics); Machine learning; Engineering","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001852135,0.0006215546,0.0004291587,0.001296575,0.0006612873,0.001939703,0.0008672328,0.002055434,0.008042742],"category_scores_gemma":[0.0233853,0.000297394,0.000362413,0.0009030539,0.001859699,0.00361954,0.001305559,0.002003459,0.003171875],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006955598,"about_ca_system_score_gemma":0.0003726755,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0007979916,"about_ca_topic_score_gemma":0.001318654,"domain_scores_codex":[0.9988759,0.0002236199,0.00008638516,0.0002614361,0.0004596571,0.00009305929],"domain_scores_gemma":[0.9901758,0.004254539,0.00214601,0.001834636,0.001328746,0.0002601605],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.004316715,0.0002843279,0.04253759,0.001047762,0.0005001239,0.004178194,0.001927169,0.07450388,0.09060349,0.06321158,0.134192,0.5826972],"study_design_scores_gemma":[0.0001899591,0.001167108,0.02022758,0.0006778078,0.0003088388,0.007145657,0.001239354,0.4908049,0.2810907,0.1135647,0.08340546,0.0001779729],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5749725,0.005434625,0.3368069,0.03363823,0.007395419,0.0003297155,0.002757477,0.00987242,0.02879271],"genre_scores_gemma":[0.9506882,0.0007023156,0.02935865,0.004134055,0.0006889257,0.00006579737,0.0009105531,0.0006574627,0.01279408],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.008042742,"threshold_uncertainty_score":0.02690572,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05328125916906568,"score_gpt":0.1795337282801119,"score_spread":0.1262524691110463,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}