์ „์ฒด ๊ธ€ 202

์ ๋Œ€์  ๊ณต๊ฒฉ ๊ฐœ๋… ๋ฐ ์œ ํ˜•

๐Ÿ’ฌ ๊ฐœ์ธ์ด ์—ฌ๋Ÿฌ ์ž๋ฃŒ๋ฅผ ์ฐธ๊ณ ํ•˜๋ฉด์„œ ์ดํ•ดํ•œ๋Œ€๋กœ ์ •๋ฆฌํ•œ ๋‚ด์šฉ์ž…๋‹ˆ๋‹ค. ์˜คํƒ€๋‚˜ ์˜ณ์ง€ ๋ชปํ•œ ์ •๋ณด๊ฐ€ ์žˆ์„ ๊ฒฝ์šฐ ๋Œ“๊ธ€๋กœ ์•Œ๋ ค์ฃผ์„ธ์š” ๐Ÿ™‚ Adversarial Attack | ์ ๋Œ€์  ๊ณต๊ฒฉ ๊ตฌ๊ธ€๋งํ•ด๋ณด๋‹ˆ๊นŒ ์ ๋Œ€์  ๊ณต๊ฒฉ์˜ ๊ฐœ๋…์„ ํฌ๊ฒŒ ๋‘ ๊ฐ€์ง€๋กœ ์ •์˜ํ•  ์ˆ˜ ์žˆ๋Š” ๊ฒƒ ๊ฐ™๋‹ค. ๋‘ ๊ฐœ ๋‹ค ๋งํ•˜๊ณ ์ž ํ•˜๋Š” ๋ฐ”๋Š” ๋™์ผํ•œ ๊ฒƒ์œผ๋กœ ์ดํ•ดํ–ˆ๊ณ  ํ‘œํ˜„์˜ ์ฐจ์ด๊ฐ€ ์•„๋‹๊นŒ ์‹ถ๋‹ค. 1. ๋”ฅ๋Ÿฌ๋‹ ๋ชจ๋ธ์ด ์žˆ์„ ๋•Œ ์ฃผ์–ด์ง„ ๋ฐ์ดํ„ฐ์— Adversarial perurbation์„ ์ ์šฉํ•˜์—ฌ Adversarial example์„ ์ƒ์„ฑํ•˜๊ณ , ๋ชจ๋ธ์ด Adversarial example์— ๋Œ€ํ•ด ์˜ค๋ถ„๋ฅ˜๋ฅผ ์ผ์œผํ‚ค๊ฒŒ ํ•˜๋Š” ๊ณต๊ฒฉ ๋ฐฉ๋ฒ•. (KISA REPORT, ์ •๋ณดํ†ต์‹ ๊ธฐํšํ‰๊ฐ€์› ์ ๋Œ€์  ๋จธ์‹ ๋Ÿฌ๋‹ ๊ธฐ์ˆ  ๋™ํ–ฅ์—์„œ ์ผ๋ถ€ ๋ฐœ์ทŒ) 2. ๋จธ์‹ ๋Ÿฌ๋‹ ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์ž์ฒด์˜ ์ทจ์•ฝ์ ์— ์˜ํ•ด ์ ๋Œ€์  ํ™˜๊ฒฝ์—์„œ ..

2021๋…„ ํšŒ๊ณ ๋ก | ์˜ฌํ•ด์˜ ํ‚ค์›Œ๋“œ

12์›” 31์ผ์— 2021๋…„์„ ๋Œ์•„๋ณด๋Š” ์‹œ๊ฐ„์„ ๊ฐ€์ง€๋ ค๊ณ  ํ–ˆ๋Š”๋ฐ ์–ด์ฉŒ๋‹ค๋ณด๋‹ˆ ํ•ด๊ฐ€ ๋ฐ”๋€ ํ›„ ์ ๋Š”๋‹ค. 1๋…„์€ ๋Š˜ ๊ฐ™์€ ์—ด๋‘๋‹ฌ์ธ๋ฐ ํ•œ ์‚ด์”ฉ ๋จน์„์ˆ˜๋ก ๋” ๋นจ๋ฆฌ ์ง€๋‚˜๊ฐ€๋Š” ๊ฒƒ ๊ฐ™๋‹ค. ๐Ÿช„ ๋ณ€ํ™” ํ•ญ์ƒ ๋น„์Šทํ•œ ์ผ๊ณผ๋ฅผ ๋ณด๋‚ด๋˜ ๋‚˜์—๊ฒŒ 2021๋…„์€ ๊ต‰์žฅํ•œ ๋ณ€ํ™”์˜ ์‹œ๊ธฐ์˜€๋‹ค. ์—ฐ์ดˆ์— ๋ณธ๊ต ์—ฐ๊ตฌ์‹ค์„ ๋“ค์–ด๊ฐ”๊ณ  ์งง๊ฒŒ๋‚˜๋งˆ ํ”„๋กœ์ ํŠธ๋ฅผ ๋งก์•„๋ดค๋‹ค. ํ•œ์ฐฝ ๋Œ€ํ•™์›๊ณผ ์ทจ์—… ์ค€๋น„ ์ค‘์—์„œ ์–ด๋–ค ๊ฑธ ์„ ํƒํ•ด์•ผํ• ์ง€ ๊ณ ๋ฏผํ•˜๋˜ ์‹œ๊ธฐ์— ๋“ค์–ด๊ฐ„ ์—ฐ๊ตฌ์‹ค์€ ์ข‹์€ ๊ฒฝํ—˜์ด์—ˆ๋‹ค. 5~6๊ฐœ์›” ์ •๋„ ์žˆ์œผ๋ฉด์„œ ํ”„๋กœ์ ํŠธ์— ์—„์ฒญ๋‚œ ์ง„์ „์„ ๊ฐ€์ ธ์˜จ ๊ฒƒ์€ ์•„๋‹ˆ์—ˆ๊ธฐ ๋•Œ๋ฌธ์— ์—ฐ๊ตฌ์‹ค์—์„œ๋Š” ์ด๋Ÿฐ ๊ฑธ ํ•˜๋Š”๊ตฌ๋‚˜ ๋ง›๋ณธ ์ •๋„์ง€๋งŒ ์ง„๋กœ ๊ฒฐ์ •์— ํฐ ๋„์›€์ด ๋˜์—ˆ๋‹ค. 4์›”๋ถ€ํ„ฐ ์ง€๊ธˆ์˜ ์—ฐ๊ตฌ์‹ค๋กœ ์ปจํƒ์„ ํ–ˆ๋‹ค. ๋ณธ๊ฒฉ์ ์ธ ์ปจํƒ์€ ์ฒ˜์Œ์ด์–ด์„œ ๋ฉ”์ผ ํ•˜๋‚˜ ์“ธ ๋•Œ๋„ ํ˜น์‹œ๋‚˜ ์‹ค์ˆ˜ํ• ๊นŒ๋ด ์ผ๋‹ค ์ง€์šฐ๊ธฐ๋ฅผ ๋ช‡ ๋ฒˆ์ด๋‚˜ ํ–ˆ๋Š”์ง€ ๐Ÿ™ƒ ๊ต์ˆ˜..

[ART] ART for TensorFlow v2 - Callable ์ฝ”๋“œ ๋ถ„์„

โœ… ์ฝ”๋“œ : https://github.com/Trusted-AI/adversarial-robustness-toolbox/blob/main/notebooks/art-for-tensorflow-v2-callable.ipynb GitHub - Trusted-AI/adversarial-robustness-toolbox: Adversarial Robustness Toolbox (ART) - Python Library for Machine Learning S Adversarial Robustness Toolbox (ART) - Python Library for Machine Learning Security - Evasion, Poisoning, Extraction, Inference - Red and Blue T..

[ART] ART for TensorFlow v2 - Keras API ์ฝ”๋“œ ๋ถ„์„

โœ… ์ฝ”๋“œ : https://github.com/Trusted-AI/adversarial-robustness-toolbox/blob/main/notebooks/art-for-tensorflow-v2-keras.ipynb GitHub - Trusted-AI/adversarial-robustness-toolbox: Adversarial Robustness Toolbox (ART) - Python Library for Machine Learning S Adversarial Robustness Toolbox (ART) - Python Library for Machine Learning Security - Evasion, Poisoning, Extraction, Inference - Red and Blue Team..

Adversarial Robustness Toolbox(ART) ์„ค์น˜

์†Œ๊ฐœ ART๋Š” IBM ์‚ฌ์—์„œ ๊ฐœ๋ฐœํ•œ ๋จธ์‹ ๋Ÿฌ๋‹ ๋ณด์•ˆ์šฉ ํŒŒ์ด์ฌ ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๋‹ค. ๊ณต์‹ ํ™ˆํŽ˜์ด์ง€์— 'IBM moved ART to LF AI in July 2020.' ๋ผ๊ณ  ๋˜์–ด์žˆ๋Š” ๊ฒƒ์„ ๋ณด๋ฉด ํ˜„์žฌ๋Š” LF AI๋ผ๋Š” ๊ณณ์—์„œ ๊ด€๋ฆฌ๋ฅผ ํ•˜๋Š” ๊ฒƒ ๊ฐ™๋‹ค. ๊ด€๋ จ ์‚ฌ์ดํŠธ ๊ณต์‹ ์‚ฌ์ดํŠธ : https://adversarial-robustness-toolbox.org/ ๊ณต์‹ docs : https://adversarial-robustness-toolbox.readthedocs.io/en/latest/ ๊ณต์‹ github : https://github.com/Trusted-AI/adversarial-robustness-toolbox ์„ค์น˜ ๊ณต์‹ ๊นƒํ—ˆ๋ธŒ๋ฅผ ์ฐธ๊ณ ํ•˜์—ฌ ์„ค์น˜๋ฅผ ์ง„ํ–‰ํ•˜๋ฉด ๋˜๊ณ  pip๋ฅผ ์ด์šฉํ•œ ๋ฐฉ๋ฒ•๊ณผ git clone์„ ์ด์šฉํ•œ ๋ฐฉ๋ฒ•..

2021-2ํ•™๊ธฐ ๋ชจ๋ฐ”์ผํ”„๋กœ๊ทธ๋ž˜๋ฐ ๊ธฐ๋ง ํ”„๋กœ์ ํŠธ

โœ ์•ˆ๋“œ๋กœ์ด๋“œ ์ŠคํŠœ๋””์˜ค ๋ง›๋ณด๊ธฐ ์™„! ์ด๋ฒˆ ํ•™๊ธฐ ์ˆ˜๊ฐ• ๊ณผ๋ชฉ ์ค‘ ๊ฐ€์žฅ ์‹ ๊ฒฝ ์“ฐ์˜€๊ณ  ๊ทธ๋งŒํผ ์‹œ๊ฐ„๋„ ๊ฝค๋‚˜ ํˆฌ์žํ–ˆ๋˜ ๊ณผ๋ชฉ์˜ ์„ฑ์ ์ด ๋‚˜์™”๋‹ค. ํ”ผ์น˜ ๋ชปํ•  ์‚ฌ์ •์œผ๋กœ ์ˆ˜๊ฐ• ์‹ ์ฒญ์„ ๊ต‰์žฅํžˆ ๋Šฆ๊ฒŒ ํ•œ ๊ณผ๋ชฉ์ด๋ผ ์ˆ˜์—…์€ 4์ฃผ์ฐจ์˜€๋‚˜ 5์ฃผ์ฐจ๋ถ€ํ„ฐ ๋“ค์€ ๋ฐ๋‹ค๊ฐ€, ๊ณผ์ œ๋„ ๊ธฐํ•œ์ด ์ง€๋‚˜ ๋ชป ๋‚ธ ๊ฒŒ ์žˆ์—ˆ๊ธฐ ๋•Œ๋ฌธ์— ํ•™์ ์— ๋Œ€ํ•œ ๊ธฐ๋Œ€์น˜๊ฐ€ ๋†’์ง€ ์•Š์•˜๋‹ค. ํ•˜์ง€๋งŒ ๊ต์ˆ˜๋‹˜๊ป˜์„œ ๊ธฐ๋ง๊ณ ์‚ฌ ๋Œ€์ฒด์˜€๋˜ ํ”„๋กœ์ ํŠธ๋ฅผ ์ข‹๊ฒŒ ๋ด์ฃผ์‹  ๊ฒƒ ๊ฐ™์•„ ๊ธฐ๋ถ„์ด ์ข‹๋‹ค. ํ•ญ์ƒ ์•ฑ ๊ฐœ๋ฐœ ๊ณต๋ถ€๋ฅผ ํ•ด๋ณด๊ณ  ์‹ถ๋‹ค๋Š” ์ƒ๊ฐ์€ ํ–ˆ์ง€๋งŒ ๋งˆ์Œ ๋จน์€ ์ผ์„ ์‹คํ–‰ํ•˜๋Š” ๋ฐ์— ์ฒœ๋งŒ ๋…„์ด ๊ฑธ๋ฆฌ๋Š” ๋‚ด ์„ฑ๊ฒฉ์ƒ ์—ญ์‹œ๋‚˜ ์•ˆํ•˜๊ณ  ์žˆ๋˜ ์™€์ค‘ ์ด๋ฒˆ ๊ณผ์ œ๊ฐ€ ์—„์ฒญ ๋„์›€์ด ๋๋‹ค. ์•ˆ๋“œ๋กœ์ด๋“œ ์ŠคํŠœ๋””์˜ค๋ฅผ ์ด์šฉํ•ด CRUD๊ฐ€ ๊ฐ€๋Šฅํ•œ ์•ฑ์„ ๋งŒ๋“œ๋Š” ๊ฒŒ ์ฃผ์ œ์˜€๋Š”๋ฐ ์•ฑ์— ๋Œ€ํ•œ ์ž์„ธํ•œ ๋‚ด์šฉ์€ ๋”ฐ๋กœ ๊ธ€์„ ์“ธ ์˜ˆ์ •! ํ•œ 2์ฃผ ์ •๋„ ์—ฐ๊ตฌ์‹ค์—์„œ ๋งจ๋‚  ๊ธฐ..

Epilogue 2021.12.27

2021/11์›”ํ˜ธ

๐Ÿ“– ํŠน๋ณ„ํ•œ ์ผ์ด ์—†์–ด๋„ ํฌ๋ฆฌ์Šค๋งˆ์Šค๋ผ๋Š” ๋‚  ํ•˜๋‚˜๋งŒ์œผ๋กœ ๊ธฐ๋ถ„์ด ์ข‹์•„์ง€๋Š” 12์›”์ด๋‹ค. ์—ฐ๊ตฌ์‹ค์—์„œ ์ง„ํ–‰ํ•œ 2๋‹ฌ ์ •๋„์˜ OJT๋„ ๋ชจ๋‘ ๋๋‚ฌ๊ณ  ์กธ์—… ๊ณผ์ œ๋„ ๋๋‚ฌ๋‹ค. ์ต์ŠคํŠธ๋ฆผ ๊ทธ ์ž์ฒด์˜€๋˜ ์กธ์—…์‚ฌ์ •์˜ ์ตœ์ข… ๊ฒฐ๊ณผ๊ฐ€ ์กธ์—…์˜ˆ์ •์ž๋กœ ๋–ด๊ธฐ ๋•Œ๋ฌธ์— ํ•™๋ถ€์ƒ์œผ๋กœ ๋ณด๋‚ด๋Š” ๋งˆ์ง€๋ง‰ ํ•™๊ธฐ์—์„œ ์œ ์ข…์˜ ๋ฏธ๋ฅผ ๊ฑฐ๋‘ฌ๋ณด๋ ค๊ณ  ํ•œ๋‹ค. ๊ธฐ๋ง๊ณ ์‚ฌ๋„ ํ…€ํ”„๋กœ์ ํŠธ๋„ ์–ด๋Š ํ•˜๋‚˜ ๋†“์น˜์ง€ ๋ง๊ณ  ์ตœ์„ ์„ ๋‹คํ•  ๊ฒƒ! โœ ๋ธ”๋กœ๊ทธ๋ฅผ ํฌ๊ฒŒ ํ•œ ๋ฒˆ ์† ๋ณด๊ณ  ์‹ถ์€๋ฐ ์–ด๋””์„œ๋ถ€ํ„ฐ ์†์„ ๋Œ€์•ผ ํ•˜๋‚˜ ์ƒ๊ฐ ์ค‘์ด๋‹ค. ์—ฐ๊ตฌ์‹ค์—์„œ ๊ณต๋ถ€ํ•œ ๊ฒƒ๋“ค๋„ ์ฐจ๊ทผ์ฐจ๊ทผ ์˜ฎ๊ฒจ๋†“์œผ๋ ค๊ณ  ํ•œ๋‹ค. ์ผ๋‹จ ์นดํ…Œ๊ณ ๋ฆฌ ์ •๋ฆฌ๋ฅผ ๋จผ์ € ํ•ด์•ผ๊ฒ ๋‹ค. ๋ธ”๋กœ๊ทธ์— ํฌํŠธํด๋ฆฌ์˜ค์Šค๋Ÿฌ์šด ๋Š๋‚Œ์„ ์ฃผ๊ณ  ์‹ถ์–ด์„œ ์Šคํ‚จ๋„ ๋œฏ์–ด ๊ณ ์น˜๊ณ  ์‹ถ์€๋ฐ ํ‹ฐ์Šคํ† ๋ฆฌ ์Šคํ‚จ ๊พธ๋ฏธ๋Š” ๊ฒŒ ๋„ค์ด๋ฒ„ ๋ธ”๋กœ๊ทธ ์Šคํ‚จ์ฒ˜๋Ÿผ ๋š๋”ฑ ํ•  ์ˆ˜ ์žˆ๋Š” ๊ฒŒ ์•„๋‹ˆ๊ธฐ ๋•Œ๋ฌธ์— ,, ์Šคํ‚จ ๊ต์ฒด๋Š” ์‚ฌ์‹ค ์–ธ์ œ ํ• ..

[๊ฐœ๋…] object detection ๊ด€๋ จ ์šฉ์–ด ์ •๋ฆฌ (1)

Classification / Localization / Object Detection Classification - ์ž…๋ ฅ์œผ๋กœ ๋“ค์–ด์˜จ ํ•˜๋‚˜์˜ ์ด๋ฏธ์ง€์— ๋Œ€ํ•ด ํ•ด๋‹นํ•˜๋Š” label ์ถœ๋ ฅ - single object๋ฅผ ๋‹ค๋ฃธ Localization - ์ž…๋ ฅ์œผ๋กœ ๋“ค์–ด์˜จ ํ•˜๋‚˜์˜ ์ด๋ฏธ์ง€์—์„œ ํ•œ object์˜ ์œ„์น˜(์ขŒํ‘œ) ์ถœ๋ ฅ - single object๋ฅผ ๋‹ค๋ฃธ - object๊ฐ€ ์žˆ๋Š” ๊ณณ์— bounding box๋ฅผ ๊ทธ๋ฆฌ๋Š” ๋ฐฉ์‹ Object Detection - ์ž…๋ ฅ์œผ๋กœ ๋“ค์–ด์˜จ ํ•˜๋‚˜์˜ ์ด๋ฏธ์ง€์—์„œ ์—ฌ๋Ÿฌ object๋ฅผ ๋ถ„๋ฅ˜ํ•˜๊ณ  ๊ฐ object๋“ค์˜ ์œ„์น˜ ์˜ˆ์ธก - multi object๋ฅผ ๋‹ค๋ฃธ - object detection์˜ ๋ฌธ์ œ ํ•ด๊ฒฐ ๋ฐฉ๋ฒ•์œผ๋กœ sliding window๊ฐ€ ๋งŽ์ด ์‚ฌ์šฉ๋˜์—ˆ์Œ IoU IoU(Intersection..

2021/10์›”ํ˜ธ

๐Ÿ“– ์—ฐ๊ตฌ์‹ค์—์„œ ๊ฐ์ฒด ํƒ์ง€ ๊ด€๋ จ ๋…ผ๋ฌธ์œผ๋กœ OJT๋ฅผ ์ง„ํ–‰ํ•˜๊ณ  ์žˆ์–ด์„œ ํ•œ ์ฃผ์˜ ๋Œ€๋ถ€๋ถ„์˜ ์‹œ๊ฐ„์„ ๋…ผ๋ฌธ ๋ถ„์„๊ณผ ์ž๋ฃŒ ๋งŒ๋“œ๋Š” ๋ฐ์— ์“ฐ๊ณ  ์žˆ๋‹ค. ์•„์ง ๋…ผ๋ฌธ ์ฝ๋Š” ์Šคํ‚ฌ๋„ ์—†๊ณ , ํ•ต์‹ฌ์ ์ธ ๋‚ด์šฉ๋งŒ ํ•„ํ„ฐ๋งํ•˜๋Š” ๋Šฅ๋ ฅ๋„ ๋ถ€์กฑํ•ด์„œ ๋…ผ๋ฌธ ํ•œ ํŽธ์„ ์ฝ๋Š” ๊ฒƒ๋„ ์‹œ๊ฐ„์ด ๊ฝค ๊ฑธ๋ฆฐ๋‹ค. ํ•˜์ง€๋งŒ ์—ฌ๋Ÿฌ ํŽธ ์ฝ๋‹ค ๋ณด๋ฉด ์ฐจ์ฐจ ๋Š˜ ๊ฑฐ๋ผ๋Š” ์ƒ๊ฐ์„ ๊ฐ€์ง€๊ธฐ๋กœ ํ–ˆ๋‹ค! ์ด ๊ธ€ ๋‹ค ์“ฐ๋ฉด ๋…ผ๋ฌธ 2๊ฐœ ์ฝ์–ด์•ผ ํ•˜๋Š”๋ฐ ๋ˆˆ์— ํž˜ ๋นก ๐Ÿ”ฅ ์ข…๊ฐ•์ด ํ•œ ๋‹ฌ ๋ฐ˜ ์ •๋„ ๋‚จ์•˜๋‹ค. ์ค‘๊ฐ„๊ณ ์‚ฌ๋„ ๋๋‚ฌ๊ณ  ์ˆ˜์—…๋„ ๋ฐ€๋ฆฌ๋Š” ๊ฒƒ ์—†์ด ์ž˜ ์ฑ™๊ธฐ๊ณ  ์žˆ๋‹ค. ํ•œ ๊ฐ€์ง€ ๋ฐ˜์„ฑํ•  ์ ์€ OCU๋กœ ๋“ฃ๊ณ  ์žˆ๋Š” ๊ณผ๋ชฉ์ด ์˜์–ด ๊ด€๋ จ๋œ ๊ณผ๋ชฉ์ธ๋ฐ ๊ณต๋ถ€๋„ ์•ˆํ–ˆ์œผ๋ฉด์„œ ๊ทผ์ž๊ฐ๋งŒ ๊ฐ€์ง€๊ณ  ์ณค๋”๋‹ˆ ์ •๋ง ๋งˆ์Œ์— ์•ˆ ๋“œ๋Š” ์ ์ˆ˜๋ฅผ ๋ฐ›์•˜๋‹ค๋Š” ๊ฒƒ์ด๋‹ค. ์–ด๋–ค ์˜์–ด ์‹œํ—˜์ด๋“  ๋‚ด ๋งˆ์Œ์— ์•ˆ ๋“œ๋Š” ์„ฑ์ ์„ ๋ฐ›์•„๋ณธ ์ ์ด ๊ฑฐ์˜ ์—†์—ˆ์–ด์„œ ๊ฑฐ๋งŒํ•œ ๋งˆ์Œ๊ฐ€์ง์„..

2021/9์›”ํ˜ธ

์–ด์ฉŒ๋‹ค ๋ณด๋‹ˆ 9์›” ์ •๋ฆฌ๋ฅผ ์ด์ œ ํ•˜๊ฒŒ ๋˜์—ˆ๋‹ค. ์ด๋ฒˆ์—๋Š” ์นดํ…Œ๊ณ ๋ฆฌ ์•ˆ ๋‚˜๋ˆ„๊ณ  ๊ทธ๋ƒฅ ์ ์–ด์•ผ์ง€ ^_^ ์–ด๋Š์ƒˆ ์—ฐ๊ตฌ์‹ค ์ถœ๊ทผํ•œ์ง€๋„ ํ•œ ๋‹ฌ์ด ๋„˜์—ˆ๋‹ค. ์•„์ง๋„ ๊น๋‘๊ธฐ ๋ผ์ดํ”„๋ฅผ ์‚ด๊ณ  ์žˆ์ง€๋งŒ ์„ ํ˜•๋Œ€์ˆ˜ ์Šคํ„ฐ๋””๋„ ํ•˜๊ณ  ๋จธ์‹ ๋Ÿฌ๋‹ ๊ฐ•์˜๋„ ๋“ฃ๊ณ  ๋…ผ๋ฌธ ์Šคํ„ฐ๋””๋„ ์ฐธ์—ฌํ•˜๊ณ  ์ด๊ฒƒ์ €๊ฒƒ ๊ฒฝํ—˜ํ•œ ์‹œ๊ฐ„์ด์—ˆ๋‹ค. 9์›” ๋™์•ˆ ๊ฐ€์žฅ ๋งŽ์ด ํ–ˆ๋˜ ์ƒ๊ฐ์€ '์–ด๋–ป๊ฒŒ ๊ณต๋ถ€ํ•ด์•ผ ์ธ๊ณต์ง€๋Šฅ ๊ณต๋ถ€๋ฅผ ์ž˜ ํ•˜๊ณ  ์žˆ๋‹ค๊ณ  ํ•  ์ˆ˜ ์žˆ์„๊นŒ?' ์˜€๋‹ค. ์—ฐ๊ตฌ์ง์œผ๋กœ ์ง„๋กœ๋ฅผ ๊ฒฐ์ •ํ•œ ๊ฑด ์•„๋‹Œ๋ฐ ๋‚ด๋…„๋ถ€ํ„ฐ 2๋…„ ๊ฐ„์€ ์—ฐ๊ตฌ์ž๋กœ ์‹œ๊ฐ„์„ ๋ณด๋‚ผ ๊ฑฐ๋‹ˆ๊นŒ ์ด๋ก ์— ์ข€ ๋” ์ดˆ์ ์„ ๋‘ฌ์•ผ ํ•˜๋‚˜ ์‹ถ๋‹ค๊ฐ€๋„ ๊ถ๊ทน์ ์œผ๋กœ๋Š” 60์‚ด์—๋„ ์ฝ”๋“œ ์งœ๋Š” ์‚ฌ๋žŒ์ด ๋˜๊ณ  ์‹ถ์€๋ฐ ์ด๊ฒŒ ๋งž๋‚˜? ์‹ถ๊ธฐ๋„ ํ•˜๋‹ค. ๋ฌผ๋ก  ์ด๋ก ์ ์ธ ๊ณต๋ถ€(์ˆ˜์‹ ํŒŒํ—ค์น˜๊ธฐ, ๋…ผ๋ฌธ ์ฝ๊ธฐ ๋“ฑ)์™€ ์ฝ”๋“œ ๋‹ค๋ฃจ๋Š” ๊ฒƒ ๋‘˜ ๋‹ค ์ž˜ํ•ด์•ผ๊ฒ ์ง€๋งŒ! ๊ฐ€๊ณ  ์‹ถ์€ ํšŒ์‚ฌ๋„ ์žˆ๊ณ  ์„์‚ฌ ์ƒํ™œํ•˜๋ฉด..