{"id":113971,"date":"2024-02-29T18:25:41","date_gmt":"2024-02-29T16:25:42","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/v-repozitorii-hugging-face-vyyavleny-vredonosnye-ai-modeli-vypolnyayushhie-kod"},"modified":"2024-02-29T18:25:41","modified_gmt":"2024-02-29T16:25:42","slug":"v-repozitorii-hugging-face-vyyavleny-vredonosnye-ai-modeli-vypolnyayushhie-kod","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/news\/v-repozitorii-hugging-face-vyyavleny-vredonosnye-ai-modeli-vypolnyayushhie-kod","title":{"rendered":"N\u00eb depozit\u00ebn Hugging Face u zbuluan modele AI dashakeqe q\u00eb ekzekutojn\u00eb kod","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
K\u00ebrkuesit nga kompania JFrog zbuluan n\u00eb depot e Hugging Face modele t\u00eb d\u00ebmshme t\u00eb m\u00ebsimit makina, instalimi i t\u00eb cilave mund t\u00eb \u00e7oj\u00eb n\u00eb ekzekutimin e kodit nga sulmuesi p\u00ebr t\u00eb marr\u00eb kontrollin mbi sistemin e p\u00ebrdoruesit. Problemi shkaktohet nga fakti q\u00eb disa formate shp\u00ebrndarjeje t\u00eb modeleve lejojn\u00eb p\u00ebrfshirjen e kodit ekzekutues, p\u00ebr shembull, modelet q\u00eb p\u00ebrdorin formatin \u00abpickle\u00bb mund t\u00eb p\u00ebrfshijn\u00eb objekte t\u00eb serializuar n\u00eb gjuh\u00ebn Python, si dhe kod q\u00eb ekzekutohet gjat\u00eb ngarkimit t\u00eb skedarit, nd\u00ebrsa modelet Tensorflow Keras mund t\u00eb ekzekutojn\u00eb kod p\u00ebrmes Lambda Layer. <\/p>\n
P\u00ebr t\u00eb parandaluar p\u00ebrhapjen e modeleve t\u00eb tilla dashakeqe, Hugging Face p\u00ebrdor skanim p\u00ebr t\u00eb zbuluar futjen e kodit t\u00eb serializuar, por modelet dashakeqe t\u00eb identifikuara tregojn\u00eb se kontrollet ekzistuese mund t\u00eb anashkalohen. P\u00ebr m\u00eb tep\u00ebr, n\u00eb shumic\u00ebn e rasteve Hugging Face thjesht i sh\u00ebnon modelet si t\u00eb rrezikshme, pa bllokuar qasjen n\u00eb to. N\u00eb total jan\u00eb identifikuar rreth 100 modele potencialisht dashakeqe, 95% e t\u00eb cilave jan\u00eb t\u00eb destinuara p\u00ebr p\u00ebrdorim me framework-un PyTorch, nd\u00ebrsa 5% me Tensorflow. Ndryshimet dashakeqe m\u00eb t\u00eb shpeshta jan\u00eb kapja e objekteve, krijimi i nj\u00eb reverse shell, nisja e aplikacioneve dhe shkrimi n\u00eb skedar.
<\/center> <\/p>\n
Vihet re se, duke gjykuar nga veprimet q\u00eb kryejn\u00eb, shumica e modeleve dashakeqe t\u00eb identifikuara jan\u00eb krijuar nga studiues t\u00eb siguris\u00eb q\u00eb p\u00ebrpiqen t\u00eb marrin shp\u00ebrblim p\u00ebr zbulimin e dob\u00ebsive dhe metodave p\u00ebr anashkalimin e mbrojtjes s\u00eb Hugging Face (p\u00ebr shembull, n\u00eb vend t\u00eb nj\u00eb sulmi real, modele t\u00eb tilla p\u00ebrpiqen t\u00eb nisin kalkulatorin ose t\u00eb d\u00ebrgojn\u00eb nj\u00eb k\u00ebrkes\u00eb rrjeti me informacion mbi suksesin e sulmit). Megjithat\u00eb, hasen edhe ekzemplar\u00eb q\u00eb nisin nj\u00eb reverse shell p\u00ebr t\u00eb lidhur sulmuesin me sistemin. <\/p>\n
P\u00ebr shembull, modelet \u00abballer423\/goober2\u00bb dhe \u00abstar23\/baller13\u00bb kan\u00eb si q\u00ebllim kryerjen e nj\u00eb sulmi ndaj sistemeve q\u00eb ngarkojn\u00eb skedarin e modelit n\u00eb PyTorch duke p\u00ebrdorur funksionin torch.load(). P\u00ebr t\u00eb organizuar ekzekutimin e kodit p\u00ebrdoret metoda \u00ab__reduce__\u00bb nga moduli pickle, e cila lejon t\u00eb futet \u00e7do kod Python n\u00eb procesin e deserializimit, q\u00eb kryhet gjat\u00eb ngarkimit t\u00eb modelit. \n Burimi: opennet.ru<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"
![\"N\u00eb](\"\/wp-content\/uploads\/2024\/02\/41044e0439dc5ab9dafa74966309f7e7.png\")
<\/center> <\/p>\n