ﺗﺠﺰﯾﻪ و ﺗﺤﻠﯿﻞ ﺗﻄﺒﯿﻘﯽ ﻣﻌﯿﺎرﻫﺎي ﺳﺎﺧﺘﺎري ﮔﺮاﻓﯿﮑﯽ ﺑﺮاي ﺷﻨﺎﺳﺎﯾﯽ ﻧﺎﻫﻨﺠﺎريﻫﺎ در ﺷﺒﮑﻪﻫﺎي اﺟﺘﻤﺎﻋﯽ آﻧﻼﯾﻦ

Comparative analysis of graphical structural metrics for identifying anomalies in online social networks

ﺷﺒﮑﻪﻫﺎي اﺟﺘﻤﺎﻋﯽ ﺑﻪ دﻟﯿﻞ اﺳﺘﻔﺎده وﺳﯿﻊ و ﻣﺤﺒﻮﺑﯿﺖ آنﻫﺎ در ﻣﻌﺮض ﻣﺸﮑﻼت اﻣﻨﯿﺘﯽ ﻫﺴﺘﻨﺪ. ﺑﻨﺎﺑﺮاﯾﻦ، ﺷﻨﺎﺳﺎﯾﯽ ﻓﻌﺎﻟﯿﺖﻫﺎي ﻏﯿﺮﻋﺎدي در ﺷﺒﮑﻪﻫﺎي اﺟﺘﻤﺎﻋﯽ، ﺑﻪ اﯾﻦ دﻟﯿﻞ ﮐﻪ ﮐﻤﮏ ﻣﯽﮐﻨﺪ ﺗﺎ اﻃﻼﻋﺎت ﻣﻬﻢ و ﻗﺎﺑﻞ ﺗﻮﺟﻬﯽ در ﻣﻮرد رﻓﺘﺎر ﮐﺎرﺑﺮان ﻏﯿﺮﻋﺎدي ﺑﻪدﺳﺖ آورده و آنﻫﺎ را ﺷﻨﺎﺳﺎﯾﯽ ﮐﻨﯿﻢ؛ ﻣﻮرد ﻧﯿﺎز اﺳﺖ. ﺑﻪ ﻣﻨﻈﻮر ﺗﺸﺨﯿﺺ ﻧﺎﻫﻨﺠﺎريﻫﺎ در ﺷﺒﮑﻪﻫﺎي اﺟﺘﻤﺎﻋﯽ، ﻣﺤﻘﻘﺎن ﻋﻤﺪﺗﺎً ﺑﻪ روﯾﮑﺮدﻫﺎي ﻣﺒﺘﻨﯽ ﺑﺮ رﻓﺘﺎر و ﺳﺎﺧﺘﺎر ﻣﺘﮑﯽ ﻫﺴﺘﻨﺪ. ﻣﺎ ﺑﺎ اﺳﺘﻔﺎده از ﻣﻌﺮﻓﯽ و ﺗﺠﺰﯾﻪ و ﺗﺤﻠﯿﻞ ﻣﻌﯿﺎرﻫﺎي ﻣﻬﻢ ﮔﺮاف ﺑﺮاي ﺗﺸﺨﯿﺺ ﻓﻌﺎﻟﯿﺖﻫﺎي ﻏﯿﺮﻋﺎدي، روﯾﮑﺮد ﻣﺒﺘﻨﯽ ﺑﺮ ﺳﺎﺧﺘﺎر ﮔﺮاف را ﮔﺴﺘﺮش ﻣﯽدﻫﯿﻢ. ﻣﻘﺎﯾﺴﻪ و اﺛﺮﺑﺨﺸﯽ اﻗﺪاﻣﺎت ﺑﺮ اﺳﺎس ﺳﻨﺠﺶﻫﺎي آﻣﺎري ﻣﺎﻧﻨﺪ دﻗﺖ، ﺑﺎزﺧﻮاﻧﯽ و F-Score و ﻫﻤﭽﻨﯿﻦ ﺑﺮ اﺳﺎس ﻧﻤﺮات ﻏﯿﺮ ﻋﺎدي ﻣﺤﺎﺳﺒﻪ و اراﺋﻪ ﺷﺪه اﺳﺖ. ارزﯾﺎﺑﯽ ﻧﻈﺮي و ﺗﺠﺮﺑﯽ روي ﭼﻨﺪ ﻣﺠﻤﻮﻋﻪ داده ﺑﺰرگ ﻧﺸﺎن ﻣﯽدﻫﺪ ﮐﻪ راﺑﻄﻪ ﺑﯿﻦ ﮔﺮه واﺳﻂ و ﺗﻌﺪاد ﻟﺒﻪﻫﺎ ﺑﺮاي ﺗﺸﺨﯿﺺ و رﺗﺒﻪ ﺑﻨﺪي ﺣﺪاﮐﺜﺮي ﺗﻌﺪاد ﻧﺎﻫﻨﺠﺎريﻫﺎ ﺑﻪ درﺳﺘﯽ ﮐﻤﮏ ﻣﯽﮐﻨﺪ.

Due to their widespread use and popularity, social networks are subject to fraudulent attacks and illegal activities and security issues. Therefore, identifying abnormal activities, especially in social networks, is important because it helps to get important and significant information about the behavior of abnormal users and identify them. In order to detect abnormalities in social networks, researchers mainly rely on behavioral and structure-based approaches. We extend the graph-structure-based approach by introducing and analyzing critical graph criteria for detecting abnormal activities. The comparison and effectiveness of the measures are based on statistical measurements such as accuracy, refreshment and F-Score and also based on unusual scores. Theoretical and empirical evaluation on a Several large datasets shows that the relationship between the node node and the number of edges to help determine the maximum number of abnormalities correctly helps.