
ãSieve-CrowdããªãªãŒã¹ v0.2.0 â ããŒãã«ããŒã¿å¿ååã»ãã©ã€ãã·ãŒä¿è·ãšã³ãžã³
çµ±èšããã600 ä»¶ã®è¡šããŒã¿ã®ãã¡ 96% ã¯ããã£ã 3 ã€ã®å±æ§ïŒå¹Žéœ¢ã»éµäŸ¿çªå·ã»æ§å¥ïŒã§ãç¹å®ã® 1 人ãçµã蟌ããã
ã ãããããŒã¿ãæ§é ããšå¿ååããã¢ã€ãã¢ãå®è£ ããŸããã
Sieve ã·ãªãŒãºã®æ°ããããŒã«ïŒããŒãã«ããŒã¿å¿ååã»ãã©ã€ãã·ãŒä¿è·ãšã³ãžã³ïŒã§ãã
ããŒãã«ããŒã¿(CSV / JSON / XML)ã k-å¿åæ§ãš distinct l-倿§æ§ã§å¿ååãã CLIãSieve-Crowd v0.2.0 ã§ãã
PyPI ã«å
¬éæžã¿ãªã®ã§ãpip install ã§å
¥ããŸãã
ãã®èšäºã¯ããã®ãªãªãŒã¹ã®èšé²ã§ãã
äœãäœã£ãã®ãããªããã®èšèšãªã®ãã
ãããŠã忢±ã®ãµã³ãã«è¡šïŒ600ä»¶ïŒãå®éã«éãããšãã«äœãèµ·ããã®ãããæž¬å®ããã宿°å€ãšãšãã«èšé²ããŸãã
ãªããæ®éã®åããåèå¥ã®ãªã¹ã¯ã«ãªãã®ãïŒããŒã¿ã®ã¹ããŒã¹æ§ãšäžææ§
ã衚ã¯ãè¡ãäžäººã®äººãæãããšããäºå®ã«æ°ã¥ããŸããã
ãã£ããã¯ãé »åºããæ
å ±æŒæŽ©ãæ
å ±ã»ãã¥ãªãã£ã«é¢ãããã¥ãŒã¹ã§ãã
ãã®åéã«ãããŠãæ»æåŽã¯ 1 ç®æçªç Žããã°åã¡ãé²åŸ¡åŽã¯ãã¹ãŠãå®ããªããã°ãªããªããšããé察称æ§ãæãç«ã£ãŠããŸãã
ããªãã¡å®å
šã«å®å
šãªã·ã¹ãã ã¯ååšããªãããšããã®ã¯ãããç¥ãããååã® 1 ã€ã§ãã
ããããå®å
šãªã»ãã¥ãªãã£ãæãç«ããªãïŒïŒã»ãã¥ãªãã£ãç¡æå³ïŒãšçµè«ä»ããã®ã¯è«çã®é£èºã§ãã
å®åçãªã»ãã¥ãªãã£ããã©ã€ãã·ãŒä¿è·ã¯ãããªã¹ã¯ãäžå®ã®éŸå€ä»¥äžã«ã³ã³ãããŒã«ãããããã«è¡ãããŸãã
ããã§ããŒã¿ç®¡çã«ã€ããŠããµãšèããŠã¿ãŸããã
ãšåæã«ãå®å
šãªéšåã§ã¯ãªããæ§é çã«å€ç«ããåŽããããŒã¿ãè£è¿ããŠãçºãçŽããŸããã
CSV ãªã©ã®ããŒãã«ããŒã¿ãå
¬éã»å
±æããéãæ°åãé»è©±çªå·ãšãã£ãçŽæ¥çãªèå¥åã¯ãåé€ãäŒãååŠçã«ãã£ãŠå®¹æã«æé€ã§ããŸãã
ããããããããçŽæ¥çèå¥åãåãé€ããåŸã§ãã£ãŠããäžè¬çãªå±æ§ã®çµã¿åãããæã€èå¥åã«ã¯æ³šæãæãå¿
èŠããããŸãã
屿§ã®çµã¿åããã«ããã¹ããŒã¹æ§ïŒçæ§ïŒ
ããŒã¿ããŒã¹ã«å«ãŸããã幎霢ããéµäŸ¿çªå·ããæ§å¥ããšãã£ã屿§ãæãæµ®ãã¹ãŠã¿ããšããããã¯åäœã§ã¯ç¹å®ã®å人ãæã瀺ããã®ã§ã¯ãªããäžè¬ã«æºèå¥åãšåŒã°ãããã®ã§ãã
ãããããããã®å°æ°ã®å±æ§ãçµã¿åããã£ããšããããŒã¿å
šäœã®ååžã«ãããŠç¹ç°ãªåããçããŸãã
éœåžéšã§ãã£ãŠãã幎霢ã现ããåºåããéµäŸ¿çªå·ãæ§å¥ãšæãåãããããšã§ããã®æ¡ä»¶ãæºãã人éãã°ã«ãŒãå
ã«èªåã²ãšãããããªãã±ãŒã¹ãããèªç¶ã«çºçããŸãã
ãããããŒã¿ã®ã¹ããŒã¹æ§ã«èµ·å ããäžææ§ã§ãã
ãµã³ãã«ããŒã¿ã瀺ã宿ž¬å€ïŒ96% ã®å€ç«
ãªããžããªã«å梱ããŠãããµã³ãã«è¡š people.csvïŒå š 600 ã¬ã³ãŒãïŒãçšããŠããã®æºèå¥åïŒå¹Žéœ¢ãéµäŸ¿çªå·ãæ§å¥ïŒã®çµã¿åãããå®éã«æž¬å®ãããšããã®åŸåãæ°å€ãšããŠæç¢ºã«çŸããŸãã
äœã®åŠçãæœããŠããªãåæç¶æ
ã®è¡šã«ãããŠãèšå®ãã屿§ã®çµã¿åããããã®ã°ã«ãŒãå
ã§èªåã ãã§ãããããªåç¬ã¬ã³ãŒããæ°ãããšã574 ä»¶ïŒå
šäœã® 96%ïŒãå ããŸããã
ã€ãŸããæ°åãªã©ã®æç¢ºãªå人æ
å ±ããããããåãé€ãã衚ã§ãã£ãŠããããŒã¿æ§é ã®èгç¹ããã¯ã倧åã®è¡ããã®çµã¿åãããæã€åäžã®ååšãšããŠå€ç«ããŠããç¶æ
ã«ãããŸãã
æ§é çãªçµã³ã€ããšå¿ååã®å¿ èŠæ§
å ¬éãããããŒã¿ãšå€éšã®çµ±èšæ å ±ãåçš®ãã£ã¬ã¯ããªã亀差ãããšããããããäžææ§ã®é«ãè¡ã¯ãåå¯ããããŒã¿ãªã³ã±ãŒãžã®èгç¹ããå€éšæ å ±ãšçµã³ã€ããããç¶æ ãçã¿åºããŸãã
çŽæ¥çãªååãå«ãŸããŠããªãããå®å
šã§ãããšããåæã ãã§ã¯ã屿§ã®çµã¿åãããæã€äžææ§ãååã«ã«ããŒããããšã¯ã§ããŸããã
ã ãããããè¡ãæã€åå¥ã®å€ç«ç¶æ
ãæ°åŠçã«è©äŸ¡ããäžå®æ°ä»¥äžã®åã屿§ãæã€çŸ€ãã®äžã«é©åã«å没ãããããã®ä»çµã¿ãå¿
èŠã«ãªããŸãã
sieve-crowd ã® k-å¿åæ§ãšã³ãžã³ã¯ããŸãã«ãã®æ§é ç課é¡ã«å¯ŸåŠããããã«èšèšããŸããã
èšèšåå: æšæž¬ããªãã»çŽæããªãã»æ±ºå®è«
ãã®èª²é¡ã解決ãããããå®è£ ã«ããã£ãŠä»¥äžã® 5 ã€ã®ååãå®ããŸããã
ãšã³ãžã³ã¯æšæž¬ããªã ã©ã®åãæ©åŸ®ã§ãã©ã®åãæºèå¥åããèªåæšæž¬ã§æ±ºããããšã¯ããŸããããããæ±ºããã®ã¯äººéã§ããinspect ã³ãã³ãã¯åããšã®èŠ³æž¬å€ããã³ããæç€ºããŸããã圹å²ã®å²ãåœãŠã¯ãŠãŒã¶ãŒã®è²¬ä»»ã«ãããŠæç€ºãããŸãã
çŽæã§ã¯ãªãæž¬å® å¿ååã®æ¢çŽ¢ãè¡ãã³ãŒããšã¯å®å šã«ç¬ç«ããå¥ã¢ãžã¥ãŒã«ã§ãã§ãããã£ãåºå衚ãããäžåºŠå³å¯ã«æ°ãçŽããŸããããã§ç®æšãšãã k ã l ã®åºæºãäžåã£ãŠããã°ãå®è¡ã¯æåŠããã衚ã¯äžåæžãåºãããŸããã
決å®è«çïŒDeterministicïŒ åãå ¥åãšåãèšå®ã§ããã°ããã©ãããã©ãŒã ãåãããã€ãåäœã§å®å šã«åäžã®ä»äºãçæããŸããæšæ¥åºãã衚ãšä»æ¥åºãã衚ã§ããã·ã¥å€ãå€ããç¶æ ãæé€ããŠããŸãã
å®è¡æäŸåã©ã€ãã©ãªé¶ Python ã®æšæºã©ã€ãã©ãªã®ã¿ã§å®çµãããpandas ã numpy ãã䜿ããŸãããå€éšäŸåã«ãããµãã©ã€ãã§ãŒã³ãªã¹ã¯ãæé€ããå®å šãªãªãã©ã€ã³ç°å¢ã§åäœããŸãã
ã¬ã·ãŒãã§ãã£ãŠæŒæŽ©ã§ã¯ãªã åŠçãçµãããšè©³çްãªå ±åæžãåºåãããŸãããããã«èšèŒãããã®ã¯ååãšéèšå€ã®ã¿ã§ããã»ã«ã®çããŒã¿ã¯äžåå«ãŸããªããããã¬ã·ãŒããå®å šã«å ±æã»å ¬éã§ããŸãã
ä»çµã¿ã®æŠèг
ããŒã¿ã®åå㯠5 çš®é¡ã§ãã
id(çŽæ¥èå¥å)/ qi(æºèå¥å)/ sensitive(æ©åŸ®)/ keep(ãã®ãŸãŸ)/ drop(åé€)ã
ä¿èšŒã¯ã宣èšããã qi ã®éåã«å¯ŸããŠæç«ããŸãã
ååã«ã¯ããããã 1 ã€ã®åœ¹å²ãäžããããŸãã
qi ã«æå®ãããåã¯ãåããšã«çšæããéå±€ãäžã£ãŠäžè¬åãããŸãã
numeric: æ°å€ããæå®ããå¹ ã®åæ°ãã§ãŒã³ã§ç¯å²ã©ãã«ãž
date: æ¥ä» â 幎æ â 幎 â 幎代 â å šäœ
prefix: å é ãæ®ããŠæ«å°Ÿãããã¹ã¯
category: çšæããéå±€ã«æ²¿ã£ãŠç²ã
exact: ãã®ãŸãŸããå šäœã
ã©ã®åãã©ã®æ®µéãŸã§ç²ããããã¯ãæ Œåæ¢çŽ¢ã§æ±ºããŸãã
é€å€ã§ããã¬ã³ãŒãæ°ã®äºç®å
ã§ k ãš l ãæºããããæ®µéã®çµã¿åãããæ¢ããæ
å ±ã®æå€±ãæå°ã®ãã®ããåé äœã®æ±ããŸã§å«ããŠæ±ºå®è«çã«éžã³ãŸãã
id åã¯åé€ããããä»®åã«çœ®ãæããããŸãã
ä»®å㯠p- ã« 16 æåã®å°æåè±æ°åãç¶ãã圢ã§ãHMAC-SHA256 ã®å€ã« base32 ãéãããã®ã§ãã
åãéµã§äœã£ãå
¬éç©ã®ããã ã§ã¯åãä»®åã«ãªããŸãã
çµã³ã€ãã¯å¯èœããã ãéµããªããã°å
ã®å€ã«ã¯æ»ããŸããã
æåŸã«ãè¡ã®é åºã SplitMix64 ã§ã·ã£ããã«ããŸãã
è¡é ãæ®ã£ãŠãããšãå
ã®è¡šãšåºåã®å¯Ÿå¿ãèªã¿åããŠããŸãããã§ãã
å®éã®åºåã®äžéšãã以äžã®ãšããã
幎霢㯠5 æ³å¹
ã®ç¯å²ã©ãã«ã«ãéµäŸ¿çªå·ã¯äž 2 æ¡ããã¹ã¯ãããååã¯ä»®åã«çœ®ãæãã£ãŠããŸãã
åºåããã CSV ã®äžéšïŒ
p-uh357djn5ixph6wf,20-24,15100**,F,cold
p-pxjmo7m2uuetuboq,20-24,16000**,M,noneã¬ã·ãŒã: ã»ã«ã®å€ãéã°ãªãå ±åæž
åŠçãå®äºãããšåºåãããã¬ããŒãã¯ãæ©æ¢°å¯èªãª JSON 圢åŒãšäººéåãã®ããã¹ã圢åŒã«å¯Ÿå¿ããŠããŸãã
ããã«èšé²ãããã®ã¯åŠçååŸã®ã¬ã³ãŒãæ°ãé€å€çããªã¹ã¯è©äŸ¡ãååã®äžè¬åã¬ãã«ãšãã£ãçµ±èšæ å ±ã®ã¿ã§ãããµã³ãã«ã®å é ã¬ã³ãŒãã«å«ãŸããŠããå®åãã¬ã·ãŒãå šæã«å¯ŸããŠæ€çŽ¢ãããšããããããæ°ã¯ 0 ä»¶ã§ãããã¬ã·ãŒããã©ãã ãå ±æããŠããå ããŒã¿ã®ã»ã«ã®å€ãå€éšã«æŒããããšã¯æ§é äžãããŸããã
600 ä»¶ã®ãµã³ãã«ãã¹ã
å®éã«ãèšå®å€ k=5ãl=2ãmax_suppression=0.05 ã§ã600 ä»¶ã® people.csv ãåŠçãããšãã®ã¬ã·ãŒãæç²ããã¡ãã§ãã
Records: 600 in, 573 out (27 suppressed, 4.5%)
Before: 574 of 600 records (96%) were one-of-a-kind on those columns. After: 0.
re-identification risk (given those columns): at most 20.0%, on average 12.4%
Diversity of "diagnosis": at least 2 different values in every crowd
Verification: measured again on the output table: passedåŠçå 574 ä»¶ïŒ96%ïŒã ã£ãåç¬ã¬ã³ãŒãããåŠçåŸã¯ 0 ä»¶ã«ãªããŸããã
åºæºãæºãããªãã£ã 27 ä»¶ïŒ4.5%ïŒã®ã¬ã³ãŒãã¯å®å šã«é€å€ãããæ®ãã® 573 ä»¶ãåºåãããŸããã
åèå¥ãªã¹ã¯ã¯æå€§ã§ã 20.0%ïŒ1/5ïŒã«æããããŠããŸãã
æ©åŸ®åã®å€æ§æ§ã«ã€ããŠããã©ã®çŸ€ãã«ãããŠã 2 çš®é¡ä»¥äžã®å€ãä¿æãããŠããŸãã
ç¹çãã¹ãã¯æåŸã®è¡ã§ãã
Verification: measured again on the output table: passed â åºåããã衚ãäºåŸçã«ç¬ç«æž¬å®ããåæ Œã確èªãããšããäºå®ã§ãã
ãŸããåãå
¥åããŒã¿ã«å¯ŸããŠå床å®è¡ãããšãããåºåããã CSV ãã¬ã·ãŒãããã€ãåäœã§å®å
šã«äžèŽããŸããã
ã³ãã³ããšçµäºã³ãŒã
å©çšã§ãã CLI ã³ãã³ãïŒ
pip install sieve-crowd
sieve-crowd inspect people.csv --config-out config.json
sieve-crowd keygen -o key.txt
sieve-crowd run people.csv -c config.json -o people_safe.csv --key-file key.txtinspect: åæ§é ã芳枬ãã圹å²å®çŸ©ã®ã²ãªåœ¢ãåºåããŸãã
keygen: 峿 ŒãªããŒããã·ã§ã³ã§æå·åŠçã·ãŒã¯ã¬ããããŒãçæããŸãã
run: å¿åååŠçã®æ¬äœãå®è¡ããŸãã
ããã»ã¹ã®çµäºã³ãŒãïŒ
0: æå
1: å ¥åºåãšã©ãŒ
2: èšå®ã»äœ¿ãæ¹ã®èª€ããäžæ£ãªå ¥å
3: æåŠïŒRefusedïŒ
çµäºã³ãŒã㯠4 段éã«ããŸããã
çµäºã³ãŒã 3 ã¯ãšã©ãŒã§ã¯ãããŸããã
èŠæ±ãããå¿ååã¬ãã«ã«ããŒã¿æ§é äžå°éããªãã£ããããå®å
šã®ããã«è¡šã®åºåãéãã«æ¢ãããšããçµæã衚ããŸãã
ãã® 4 ã€ã¯å
šéšå®éã«ããããŠç¢ºèªããŸããã
k ã 1 ã«ããã 2ãã©ãäžè¬åããŠãå±ããªã k 㯠3ãååšããªããã¡ã€ã«ã¯ 1 ãè¿ããŸãã
ãã«ããã©ãŒããã察å¿ãšãã¹ãäœå¶
v0.2.0 ã®ãã¹ãã¹ã€ãŒã㯠83 ä»¶(æ¢å 42 + 圢åŒã¢ããã¿ 41)ã§ãCI 㯠Python 3.10 / 3.11 / 3.12 ã®ãããªã¯ã¹ã§åã£ãŠããŸãã
v0.2.0 ã®å¯Ÿå¿åœ¢åŒã¯ CSV / JSON / XML ã® 3 çš®ã
Excel(.xlsx / .xls)ã¯æå³çã«å¯Ÿå¿ããŸãã â äŸåã©ã€ãã©ãªãŒãã貫ãããã§ããããããŸã§ã®ã·ãªãŒãºãšåæ§ã«ã鿬¡æ¡åŒµããŠããæ¹éã§ãã
ãã®ä»£ããããåãè«ç衚ã¯ãã©ã®åœ¢åŒã§éãã§ããã€ãåäžã®åºåã«ãªããããšããã¹ãã§åºå®ããŠãããŸãã
JSON ã§èªãã§ XML ã§æžãããšããã®ãã§ããŸãã
åãªãªãŒã¹ã§æéããããã®ã¯ãåçŸæ§ã®åºå®ã§ãã
ä¹±æ°(SplitMix64)ã®ãã¯ãã«ãšä»®åã®ãã¯ãã«ã tests/vectors ã«åºå®ããŸããã
å¿ååããŒã«ãéæ±ºå®è«çã ãšãæšæ¥ã®è¡šãšä»æ¥ã®è¡šã®å¯Ÿå¿ãŸã§æŒããŠããŸãã
決å®è«ã¯å¥œã¿ã§ã¯ãªããèŠä»¶ã§ãã
v0.2.0 æç¹ã§ãåèš 83 ä»¶ã®ãã¹ãã¹ã€ãŒãã GitHub Actions äžã® Python ãããªã¯ã¹ã§åžžæã°ãªãŒã³ãç¶æããŠããŸãã
æ¢ç¥ã®éç
ä¿èšŒã¯å®£èšããã QI åã«å¯Ÿããçžå¯Ÿå€ã§ããæ»æè ãç¥ãåŸããã¹ãŠã®åãæå®ãå¿ããå Žåããã®å€åŽããã®åèå¥ãªã¹ã¯ã¯æ®ããŸãã
l-倿§æ§ã¯ distinct çã§ããæ¡ä»¶ãæºãããªã矀ãããšé€å€ããã¢ãããŒãããšã£ãŠããŸãã
Excel 圢åŒã¯é察å¿ã§ãã
æ Œåæ¢çŽ¢ã«ã¯å®å šæ§ã®äžéããããŸãïŒæ¢å®ã§ã¯ 20 äžè¡ãè¶ ããå Žåãªã©ïŒã
ãããã«
ããŒã¿ãä¿ç®¡ã»é瀺ã»å
±æããŠè¯ããããšããåãã«ãæçµçãªå€æãäžãã®ã¯äººéã§ãã
ããã§ããã®å€æã®æ ¹æ ãšãªãã客芳çãªæž¬å®å€ãçšæããããšã¯ãããããŠæè¡çã«å¯èœã§ãã
Sieve-Crowd ã¯ããã®åœ¹å²åæ
ãå¿ å®ã«ã³ãŒãã«èœãšã蟌ãã ããŒã«ã§ãã
詳现ãªä»æ§ãããã¥ã¡ã³ããå°å ¥æé ã«ã€ããŠã¯ä»¥äžãã芧ãã ããã
é¢é£ãªã³ã¯
Sieve ãšã³ã·ã¹ãã
Sieveã·ãªãŒãºå šäœã®èšèšææ³ãåãšã³ãžã³ã®åœ¹å²ã芳枬空éã®é¢ä¿ã«ã€ããŠã¯ãã¢ãŒããã¯ãã£ã€ã³ããã¯ã¹ãåç §ããŠãã ããã
Sieve ãšã³ã·ã¹ãã ã¢ãŒããã¯ãã£ã€ã³ããã¯ã¹ïŒnoteïŒ
PyPI â https://pypi.org/user/neguseatama/
GitHub ãªããžããª
Sieve-Core â æ±ºå®è«çãªæ§é éçŽã»ç¢ºä¿¡åºŠè©äŸ¡ã®åºç€ãšã³ãžã³
Sieve-Referee â ããã¹ãã®æ§é ã»æµçšã»èšãæãã芳枬ãããšã³ãžã³
Sieve-Scope â ãœãŒã¹ã³ãŒãã®æ§é çé¡äŒŒæ§ã芳枬ãããšã³ãžã³
Sieve-Lens â äžå¯èŠã³ã³ãã³ããé èœèŠçŽ ã芳枬ãããšã³ãžã³
Sieve-Redact â æå®é åã®æ±ºå®è«çãã¹ãã³ã°ãšå®å šæ§æ€èšŒ
Sieve-Order â ããã³ããã®è¡šå±€æ§é ã6-bit芳枬空éã§èŠ³æž¬ãããšã³ãžã³
Sieve-Crowd â ããŒãã«ããŒã¿å¿ååã»ãã©ã€ãã·ãŒä¿è·ãšã³ãžã³
Sieve-Tray â åãšã³ãžã³ãçµ±åããããŒã«ã«ããã³ããšã³ã
note æè¡èšäº
Sieve-ScopeïŒæ±ºå®è«çã¹ã³ãŒã管çãšã³ã³ããã¹ãå¢çã®å®çŸ©
Sieve-RedactïŒæ±ºå®è«çãã¹ãã³ã°ãšä»ãã€ãå®å šæ§æ€èšŒã«ããããŒã¿ä¿å š
Sieve-TrayïŒè€æ°ã®èŠ³æž¬ç©ºéãçµ±åããããŒã«ã«æ€èšŒããã³ããšã³ã
èšèšã®èšé²(ææèšäº)
Sieve ã·ãªãŒãºã®æ°çã¢ãã« â å°åºæ³ãšå®å®æ§
ã·ãªãŒãºã®ååŒããå£ãæ¹ã®å®æž¬ãããã©ãå°ãããããã
æèšç®ã§è¿œããç²åºŠã§è§£èª¬ãå°åºã® 3 段æé ãšèšèšãã§ãã¯ãªã¹ãä»ããSieve ã·ãªãŒãºã®æ§é èšèš â äŸåæ¹åãã¢ãŒããã¯ãã£ã決ãã
åäžãã¡ã€ã«ã»ãã©ã°ã€ã³ã»ã«ã¿ãã°çé¢ã»byte pinã5 ã€ã®æ§é 倿ã®
çç±ãšåäŸããããŠãã»ãŠã¹ã®è¹ããå§ãŸã£ãéçºçµç·¯ã
Sieveã·ãªãŒãºã«å ±éããèšèšæ¹éã¯ã芳枬ã§ãã蚌æ ãæ±ºå®è«çã«èšé²ãããã®æ ¹æ ã人éã远跡ã§ããããã«ããããšã§ãã
Observer, not Arbiter â 芳枬è ã§ãã£ãŠãè£å®è ã§ã¯ãªãã