• Àüü
  • ÀüÀÚ/Àü±â
  • Åë½Å
  • ÄÄÇ»ÅÍ
´Ý±â

»çÀÌÆ®¸Ê

Loading..

Please wait....

±¹³» ³í¹®Áö

Ȩ Ȩ > ¿¬±¸¹®Çå > ±¹³» ³í¹®Áö > Çѱ¹Á¤º¸°úÇÐȸ ³í¹®Áö > Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)

Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)

Current Result Document : 7 / 397 ÀÌÀü°Ç ÀÌÀü°Ç   ´ÙÀ½°Ç ´ÙÀ½°Ç

ÇѱÛÁ¦¸ñ(Korean Title) µ¥ÀÌÅͺ£À̽º ³»ºÎÀÚ °ø°ÝŽÁö¸¦ À§ÇÑ »ç¿ëÀÚ ÁúÀÇÀÇ ºÐ¸®Ç¥Çö ÇнÀ
¿µ¹®Á¦¸ñ(English Title) (Disentangled Representation Learning of User Queries for Database Insider Attack Detection System)
ÀúÀÚ(Author) °í±¤¸í   ºÎ¼®ÁØ   Á¶¼º¹è   Gwang-Myong Go   Seok-Jun Bu   Sung-Bae Cho  
¿ø¹®¼ö·Ïó(Citation) VOL 27 NO. 02 PP. 0076 ~ 0082 (2021. 02)
Çѱ۳»¿ë
(Korean Abstract)
¿ªÇÒ±â¹Ý Á¢±Ù Á¦¾î¸¦ ±â¹ÝÀ¸·Î ÇÏ´Â µ¥ÀÌÅͺ£À̽º °ü¸® ½Ã½ºÅÛÀº Á¤º¸ ÀúÀå ¹× ºÐ¼®¿¡ ³Î¸® »ç¿ëµÇÁö¸¸ ¿©·¯ º¸¾È À̽´ Áß¿¡¼­µµ ³»ºÎÀÚ °ø°Ý¿¡ ƯÈ÷ Ãë¾àÇÏ´Ù´Â °ÍÀÌ ¿©·¯ ¿¬±¸¸¦ ÅëÇØ ¹àÇôÁ® ÀÖ´Ù. ±¸¹® ºÐ¼®À» ÅëÇÑ ÀüÅëÀûÀΠħÀÔŽÁöÀÇ ÇÑ°è·Î ÀÎÇØ ÃÖ±ÙÀÇ ¿¬±¸°á°ú´Â ÀûÀÀÇü ½Ã½ºÅÛÀ¸·Î ¿ä¾àµÉ ¼ö ÀÖÀ¸¸ç, ÀÌ·¯ÇÑ °üÁ¡¿¡¼­ ¿ì¸®ÀÇ ¿¬±¸´Â µ¥ÀÌÅͺ£À̽º¿¡ Á¢±ÙÇÏ´Â »ç¿ëÀÚ ÁúÀÇ¿¡ ´ëÇÑ ºÐ·ù¿¹ÃøÀ» ¼öÇàÇÏ¿© ½ÇÁ¦ µ¥ÀÌÅͺ£À̽º ½Ã½ºÅÛ¿¡ ÀÇÇØ ¼öÇàµÈ ³»¿ë ºñ±³¸¦ ÅëÇØ ¿¹Ãø°ª°ú »óÀÌÇÑ °æ¿ì ³»ºÎÀÚ °ø°ÝÀ¸·Î ÆÇ´ÜÇÏ´Â ¹æ¹ýÀ» Á¦¾ÈÇÑ´Ù. Á¦¾ÈÇÏ´Â ¸ðµ¨Àº »óÈ£ À¯»ç¼ºÀÌ Å« »ç¿ëÀÚ ÁúÀÇ¿¡ ´ëÇÑ ºÐ·ù¶ó´Â ¹®Á¦ ÇØ°áÀ» À§ÇØ ÀÔ·ÂÀÇ À¯ÀǹÌÇÑ Æ¯Â¡À» ¸ðÇüÀÌ Àß ÃßÃâÇÏ°í, ½Å°æ¸ÁÀ» »ç¿ëÇÏ¿© À¯»ç¼ºÀÇ Ã´µµ¸¦ Á÷Á¢ÀûÀ¸·Î ÇнÀÇÏ´Â °èÃþÀû ±¸Á¶¸¦ °¡Áö´Â ½ÉÃþ Ç¥Çö ÇнÀ ½Å°æ¸ÁÀ¸·Î, ÇнÀ¸ðµ¨Àº ¿Â¶óÀÎ °Å·¡ º¥Ä¡¸¶Å©ÀÎ TPC-E °ø°³ ¼³°è±¸Á¶¸¦ È°¿ëÇÏ¿© °¢°¢ÀÇ ¿ªÇÒ·Î ±¸ºÐµÈ 11°³ÀÇ ºÐ·ù¸ðÇü´ç 1,000°³ÀÇ ÁúÀǸ¦ »ý¼ºÇÏ¿© ÇнÀµÇ¾úÀ¸¸ç, ±âÁ¸ ¼±Ç࿬±¸¿Í ºñ±³ÇßÀ» ¶§ °¡Àå ³ôÀº ¼º´ÉÀÎ 94.17%ÀÇ ºÐ·ùÁ¤È®µµ¸¦ ´Þ¼ºÇÏ¿´´Ù. Á¦¾ÈÇÏ´Â ¹æ¹ýÀÇ Á¤·®Àû ¼º´ÉÀ» 10°ã ±³Â÷ °ËÁõÀ¸·Î Æò°¡ÇÏ¿´°í, Á¤¼ºÀû ¼º´É ºÐ¼®À» À§ÇØ ½Å°æ¸ÁÀ¸·Î ÀÓº£µùÇÑ Æ¯Â¡°ø°£À» ½Ã°¢È­ÇÏ¿© °áÇÔ°£ÀÇ ¾ÐÃà º¤ÅÍÀÇ ±ºÁýÈ­ ºÐ¼®À» ¼öÇàÇÏ¿´´Ù.
¿µ¹®³»¿ë
(English Abstract)
Database management systems employing role-based access control are widely used for information storage and analysis, but several studies have revealed that such systems are particularly vulnerable to insider attacks, among various other security issues. Our study proposes a method that can verify an event as an insider attack when the predicted value is different from the actual value by performing classification prediction on the user query accessing the database and comparing it with the log performed by the actual database system. Our model for solving the problem of classification of user queries with high mutual similarity is a deep representation learning architecture with a hierarchical structure, in which the model extracts meaningful features and directly learns the measure of similarity using a network. The model was trained by generating 1,000 queries per 11 roles classified using the TPC-E public schema as an online transaction benchmark, and it achieved higher performance than any of the previous works, with a classification accuracy of 94.17%. The quantitative performance was evaluated by 10-fold cross-validation to verify the validity of the intrusion detection model, while for qualitative performance analysis, a clustering analysis of compression vectors between defects was conducted by visualizing the embedded feature space.
Å°¿öµå(Keyword) µö·¯´×   Ç¥ÇöÇü ÇнÀ   ½ÉÃþ »ïÁßÇ× ½Å°æ¸Á   µ¥ÀÌÅͺ£À̽º º¸¾È   ³»ºÎÀÚ Ä§ÀÔŽÁö   »ç¿ëÀÚ ÁúÀÇ Æ¯Â¡ÃßÃâ   deep learning   metric learning   triplet network   database security   insider intrusion detection   user query feature extraction  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå