{"id":103157,"date":"2022-02-01T15:36:44","date_gmt":"2022-02-01T13:36:46","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/metod-identifikaczii-sistemy-polzovatelya-na-osnove-informaczii-o-gpu"},"modified":"2022-02-01T15:36:44","modified_gmt":"2022-02-01T13:36:46","slug":"metod-identifikaczii-sistemy-polzovatelya-na-osnove-informaczii-o-gpu","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/news\/metod-identifikaczii-sistemy-polzovatelya-na-osnove-informaczii-o-gpu","title":{"rendered":"Metoda e identifikimit t\u00eb sistemit t\u00eb p\u00ebrdoruesit bazuar n\u00eb informacionin mbi GPU","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
Hulumtuesit nga Universiteti Ben-Gurion (Izrael), Universiteti i Lille (Franc\u00eb) dhe Universiteti i Adelaide ( Australi) zhvilluan nj\u00eb teknik\u00eb t\u00eb re p\u00ebr identifikimin e pajisjeve t\u00eb p\u00ebrdoruesve p\u00ebrmes p\u00ebrcaktimit t\u00eb parametrave t\u00eb pun\u00ebs s\u00eb GPU n\u00eb shfletuesin web. Metoda mori emrin 'Drawn Apart' dhe bazohet n\u00eb p\u00ebrdorimin e WebGL p\u00ebr t\u00eb marr\u00eb nj\u00eb profil t\u00eb performanc\u00ebs s\u00eb GPU, i cili mund\u00ebson nj\u00eb sakt\u00ebsi t\u00eb lart\u00eb n\u00eb metodat pasive t\u00eb gjurmimit, t\u00eb cilat funksionojn\u00eb pa p\u00ebrdorimin e cookie-ve dhe pa ruajtjen e identifikuesve n\u00eb sistemin e p\u00ebrdoruesit. <\/p>\n
Metodat q\u00eb marrin parasysh ve\u00e7orit\u00eb e caktimit, GPU-s\u00eb, grafik\u00ebs dhe drejtuesve jan\u00eb p\u00ebrdorur edhe m\u00eb par\u00eb, por ato ishin t\u00eb kufizuara n\u00eb aft\u00ebsin\u00eb p\u00ebr t\u00eb ndar\u00eb pajisjet vet\u00ebm n\u00eb nivel t\u00eb modeleve t\u00eb ndryshme t\u00eb kartave grafike dhe GPU-s\u00eb, duke qen\u00eb se mund t\u00eb sh\u00ebrbenin vet\u00ebm si nj\u00eb faktor shtes\u00eb p\u00ebr t\u00eb rritur probabilitetin e identifikimit. Karakteristika kryesore e metod\u00ebs s\u00eb re 'Drawn Apart' \u00ebsht\u00eb se ajo nuk \u00ebsht\u00eb e kufizuar n\u00eb ndarjen e modeleve t\u00eb ndryshme t\u00eb GPU-s\u00eb, por p\u00ebrpiqet t\u00eb zbuloj\u00eb dallime mes GPU-ve identike t\u00eb nj\u00eb modeli, q\u00eb vijn\u00eb si pasoj\u00eb e heterogjenitetit t\u00eb procesit t\u00eb prodhimit t\u00eb \u00e7ip\u00ebve t\u00eb dizajnuar p\u00ebr llogaritje masive paralele. Theksohet se variacionet q\u00eb lindin gjat\u00eb procesit t\u00eb prodhimit lejojn\u00eb formimin e gjurm\u00ebve t\u00eb paidentike p\u00ebr modelet e nj\u00ebjta t\u00eb pajisjeve.
<\/center> <\/p>\n
Doli se k\u00ebto dallime mund t\u00eb identifikohen p\u00ebrmes num\u00ebrimit t\u00eb blloqeve t\u00eb ekzekutimit dhe analiz\u00ebs s\u00eb performanc\u00ebs s\u00eb GPU-s\u00eb. Si primitv\u00eb p\u00ebr t\u00eb identifikuar modelet e ndryshme t\u00eb GPU-s\u00eb, jan\u00eb p\u00ebrdorur kontrollime bazuar n\u00eb nj\u00eb grup funksionesh trigonometrike, operacione logjike dhe llogaritje me pik\u00eb flottante. P\u00ebr t\u00eb zbuluar dallimet n\u00eb GPU-t\u00eb e nj\u00ebjta, \u00ebsht\u00eb vler\u00ebsuar numri i rrjedhave q\u00eb ekzekutohen nj\u00ebkoh\u00ebsisht gjat\u00eb ekzekutimit t\u00eb shenjuesve t\u00eb kulmit. Supozohet se efekti i zbuluar \u00ebsht\u00eb i shkaktuar nga dallimet n\u00eb regjimin e temperatur\u00ebs dhe konsumit t\u00eb energjis\u00eb s\u00eb ekzemplar\u00ebve t\u00eb ndrysh\u00ebm t\u00eb \u00e7ip\u00ebve (m\u00eb par\u00eb ky efekt ishte demonstruar p\u00ebr CPU-t\u00eb - procesor\u00ebt e nj\u00ebjt\u00eb, gjat\u00eb ekzekutimit t\u00eb t\u00eb nj\u00ebjtit kod, treguan konsum t\u00eb ndrysh\u00ebm energjie). <\/p>\n
Duke qen\u00eb se operacionet p\u00ebrmes WebGL ekzekutohen n\u00eb m\u00ebnyr\u00eb asinkrone, p\u00ebr matjen e koh\u00ebs s\u00eb ekzekutimit t\u00eb tyre nuk mund t\u00eb p\u00ebrdoret direkt API i JavaScript performance.now(), prandaj jan\u00eb propozuar tre truqe p\u00ebr matjen e koh\u00ebs: <\/p>\n
\n
onscreen - vizatimi i sken\u00ebs n\u00eb HTML canvas me matjen e koh\u00ebs s\u00eb aktivizimit t\u00eb funksionit callback, i caktuar p\u00ebrmes API Window.requestAnimationFrame dhe i thirrur pas p\u00ebrfundimit t\u00eb vizatimit.\n
offscreen - p\u00ebrdorimi i nj\u00eb punonj\u00ebsi dhe vizatimi i sken\u00ebs n\u00eb objektin OffscreenCanvas me matjen e koh\u00ebs s\u00eb ekzekutimit t\u00eb komand\u00ebs convertToBlob.\n
GPU - vizatimi n\u00eb objektin OffscreenCanvas, por p\u00ebrdorimi p\u00ebr matjen e koh\u00ebs t\u00eb timerit t\u00eb ofruar n\u00eb WebGL, i cili merr parasysh koh\u00ebzgjatjen e ekzekutimit t\u00eb nj\u00eb grupi komandash n\u00eb an\u00ebn e GPU-s\u00eb. <\/ul>\n
Gjat\u00eb krijimit t\u00eb identifikatorit n\u00eb secilin pajisje kryhen 50 kontrolle, secila prej t\u00eb cilave mbulon 176 matje t\u00eb 16 karakteristikave t\u00eb ndryshme. Nj\u00eb eksperiment, n\u00eb t\u00eb cilin u mblodh informacioni p\u00ebr 2500 pajisje me 1605 GPU t\u00eb ndryshme, tregoi nj\u00eb rritje t\u00eb efikasitetit t\u00eb metodave t\u00eb kombinuara t\u00eb identifikimit me 67% me shtimin e mb\u00ebshtetjes Drawn Apart. N\u00eb ve\u00e7anti, metoda e kombinuar FP-STALKER mesatarisht ofronte identifikim p\u00ebr 17.5 dit\u00eb, nd\u00ebrsa n\u00eb kombinim me Drawn Apart, shkalla e identifikimit u rrit n\u00eb 28 dit\u00eb.
<\/center> <\/p>\n
\n
Sakt\u00ebsia e ndarjes s\u00eb 10 sistemeve me \u00e7ipa Intel i5-3470 (GEN 3 Ivy Bridge) dhe GPU Intel HD Graphics 2500 n\u00eb testin onscreen ishte 93%, nd\u00ebrsa n\u00eb offscreen - 36.3%.\n
P\u00ebr 10 sisteme Intel i5-10500 (GEN 10 Comet Lake) me kart\u00ebn grafike NVIDIA GTX1650, sakt\u00ebsia ishte 70% dhe 95.8%.\n
P\u00ebr 15 sisteme Intel i5-8500 (GEN 8 Coffee Lake) me GPU Intel UHD Graphics 630 - 42% dhe 55%.\n
P\u00ebr 23 sisteme Intel i5-4590 (GEN 4 Haswell) me GPU Intel HD Graphics 4600 - 32.7% dhe 63.7%.\n
P\u00ebr gjasht\u00eb smartphone Samsung Galaxy S20\/S20 Ultra me GPU Mali-G77 MP11 n\u00eb testin me output n\u00eb ekran, sakt\u00ebsia e identifikimit ishte 92.7%, nd\u00ebrsa p\u00ebr smartphone-t Samsung Galaxy S9\/S9+ me Mali-G72 MP18 - 54.3%. <\/ul>\n
<\/center> <\/p>\n
V\u00ebrehet se sakt\u00ebsin\u00eb e ndikuar temperatur\u00ebn e GPU-s\u00eb, dhe p\u00ebr disa pajisje rinisja e sistemit \u00e7onte n\u00eb shkarraviten e identifikuesit. Duke p\u00ebrdorur k\u00ebt\u00eb metod\u00eb n\u00eb kombinim me metoda t\u00eb tjera t\u00eb identifikimit indirekt, sakt\u00ebsia mund t\u00eb rritet ndjesh\u00ebm. Gjithashtu, planifikojn\u00eb t\u00eb rrisin sakt\u00ebsin\u00eb p\u00ebrmes p\u00ebrdorimit t\u00eb shenjuesve t\u00eb llogaritjes pas stabilizimit t\u00eb API-s\u00eb s\u00eb re WebGPU. <\/p>\n
Kompani si Intel, ARM, Google, Khronos, Mozilla dhe Brave u njoftuan p\u00ebr problemin q\u00eb n\u00eb vitin 2020, por detajet e metod\u00ebs jan\u00eb zbuluar vet\u00ebm tani. N\u00eb mesin e tyre, hulumtuesit publikuan shembuj funksional\u00eb n\u00eb JavaScript dhe GLSL, q\u00eb mund t\u00eb funksionojn\u00eb me ose pa shfaqje informacioni n\u00eb ekran. Gjithashtu, p\u00ebr sistemet me GPU Intel GEN 3\/4\/8\/10 jan\u00eb publikuar grupe t\u00eb dh\u00ebnash p\u00ebr klasifikimin e informacionit t\u00eb nxjerr\u00eb n\u00eb sistemet e m\u00ebsimit t\u00eb makinerive. \n Burimi: opennet.ru<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"
![\"Metoda](\"\/wp-content\/uploads\/2022\/02\/947fc4477ff99c14ca7854e36e4ad976.jpg\")
<\/center> <\/p>\n![\"Metoda](\"\/wp-content\/uploads\/2022\/02\/c57096ca8d5833df92cd7f992f19a289.jpg\")
<\/center> <\/p>\n![\"Metoda](\"\/wp-content\/uploads\/2022\/02\/e5462f5b761dc6e16a43ac00d8f0b7aa.jpg\")
<\/center> <\/p>\n