Àá½Ã¸¸ ±â´Ù·Á ÁÖ¼¼¿ä. ·ÎµùÁßÀÔ´Ï´Ù.

µö·¯´× ±â¹Ý ÀÇ·á ¿µ»ó ÀΰøÁö´É ¸ðµ¨ÀÇ Ãë¾à¼º: Àû´ëÀû °ø°Ý Exploiting the Vulnerability of Deep Learning-Based Artificial Intelligence Models in Medical Imaging: Adversarial Attacks

´ëÇÑ¿µ»óÀÇÇÐȸÁö 2019³â 80±Ç 2È£ p.259 ~ 273
±èÈÖ¿µ, Á¤´ëö, ÃÖº´¿í,
¼Ò¼Ó »ó¼¼Á¤º¸
±èÈÖ¿µ ( Kim Whi-Young ) 
¿¬¼¼´ëÇб³ ÀÇ°ú´ëÇÐ ÀǷ῵»óµ¥ÀÌÅÍ»çÀ̾𽺼¾ÅÍ

Á¤´ëö ( Jung Dae-Chul ) 
¿¬¼¼´ëÇб³ ÀÇ°ú´ëÇÐ ¼¼ºê¶õ½ºº´¿ø ¿µ»óÀÇÇаú
ÃÖº´¿í ( Choi Byoung-Wook ) 
¿¬¼¼´ëÇб³ ÀÇ°ú´ëÇÐ ¼¼ºê¶õ½ºº´¿ø ¿µ»óÀÇÇаú

Abstract

µö·¯´× ÇнÀ¸ðµ¨ ¼º´ÉÀÇ ºñ¾àÀûÀÎ ¹ßÀüÀ¸·Î ÀÎÇØ ¿µ»óÀÇÇÐÀ» Áß½ÉÀ¸·Î ÇÏ¿© ±â°èÇнÀ ¸ðµ¨µéÀÌ ½ÇÁ¦ ÀÓ»óÇöÀå¿¡¼­ Àǻ縦 º¸Á¶ÇÏ¿© Áø´Ü´ÉÀ» ³ôÀÌ°í ÀÛ¾÷ÀÇ È¿À²¼ºÀ» Áõ´ë ½ÃÄÑÁÙ °ÍÀ̶ó´Â ±â´ë°¡ ¸¹´Ù. ÀÌ·¯ÇÑ ±â´ë·Î ÀÎÇØ ¸¹Àº º´¿ø°ú ¹Î°£ ±â¾÷ µî¿¡¼­ ÀÇÇÐ ¿µ»óÀ» ÀÌ¿ëÇÑ ÀÚµ¿ Áø´Ü ÇнÀ¸ðµ¨ °³¹ß °æÀïÀÌ ¶ß°Ì´Ù. ½ÇÁ¦·Î °¡±î¿î ¹Ì·¡¿¡ ¸¹Àº µö·¯´× ±â¹ÝÀÇ ÀÚµ¿ Áø´Ü ÇÁ·Î±×·¥µéÀÌ ÀÇ·á ȯ°æ¿¡¼­ »ç¿ëµÉ °ÍÀÌ´Ù. ±×·¯³ª, µö·¯´× ¾Ë°í¸®µëÀÌ ³»ÀçÀûÀ¸·Î °¡Áø ºÒÈ®½Ç¼º(uncertainty)¿¡ ÀÇÇÑ Àû´ëÀû °ø°Ý(adversarial attack)ÀÇ °¡´É¼ºÀº ƯÈ÷ ÀÇÇй®Á¦¿¡ µö·¯´× ¾Ë°í¸®µëÀ» Àû¿ëÇÏ´Â µ¥¿¡ Å« °É¸²µ¹ÀÌ µÈ´Ù. º» Á¾¼³¿¡¼­´Â ÀÇÇпµ»óÀ» ´Ù·ç´Â µö·¯´× ¸ðµ¨µé¿¡ ´ëÇØ ¾î¶°ÇÑ ¿ø¸®¿Í ¹æ½ÄÀ¸·Î Àû´ëÀû °ø°ÝÀÌ ÀÌ·ç¾îÁú ¼ö ÀÖÀ¸¸ç, ÀÌ·Î ÀÎÇÏ¿© ¾î¶² ¹®Á¦µéÀÌ ¹ß»ýÇÒ ¼ö ÀÖÀ¸¸ç Àû´ëÀû °ø°ÝÀ» Â÷´ÜÇÒ ¼ö ÀÖ´Â ¹æ¹ýÀº ¾ø´ÂÁö ÀÚ¼¼È÷ »ìÆ캸°íÀÚ ÇÑ´Ù.

Due to rapid developments in the deep learning model, artificial intelligence (AI) models are expected to enhance clinical diagnostic ability and work efficiency by assisting physicians. Therefore, many hospitals and private companies are competing to develop AI-based automatic diagnostic systems using medical images. In the near future, many deep learning-based automatic diagnostic systems would be used clinically. However, the possibility of adversarial attacks exploiting certain vulnerabilities of the deep learning algorithm is a major obstacle to deploying deep learning-based systems in clinical practice. In this paper, we will examine in detail the kinds of principles and methods of adversarial attacks that can be made to deep learning models dealing with medical images, the problems that can arise, and the preventive measures that can be taken against them.

Å°¿öµå

Deep Learning; Artificial Intelligence; Medical Imaging

¿ø¹® ¹× ¸µÅ©¾Æ¿ô Á¤º¸

 

µîÀçÀú³Î Á¤º¸

KCI
KoreaMed
KAMS