Classification of Confidential Images Using Neural Hash

Humanity generates considerable information using its devices – smartphones, laptops, and tablets. Users upload images to different platforms, such as social networks, messengers, web services other applications, which greatly endanger their personal information. User privacy has been exploited on the Internet for a long time. Interested parties lure potential customers into trap of offers age, weight, nationality, religion preferences. The sensitive that may be contained in is sometimes not recognized by users dangerous share and, therefore, can easily shared online owner without second thought.This article inspects neural hash algorithm solving image classification tasks confidential evaluates it via basic metrics. main idea find similar will serve an example defining classes. uses codes, ensuring users’ privacy. evaluation based “The Visual Privacy (VISPR) Dataset”. components are network vectors extracted features indexed set (hash tables) store knowledge about particular domain.The critical aspect involves collisions codes due similarity features. resulting identical or differ specific value Hamming distance. Multiple tables with functions used increase recall precision results. effect imperfect taxonomy was analyzed, led further filtration abstract classes increasing overall scores.Also, investigates “pseudo-adaptivity” - ability classify new add cases existing were included training stages. Such crucial domains many instances