Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Sign up for free
Menu
Search
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Pricing
Search
Sign in
Sign up for free
ディープラーニングによる通信トラヒックの識別
Search
Kazutoshi Nakano
May 13, 2023
Research
56
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
ディープラーニングによる通信トラヒックの識別
ディープラーニングによる通信トラヒックの識別
Kazutoshi Nakano
May 13, 2023
More Decks by Kazutoshi Nakano
See All by Kazutoshi Nakano
Image data augmentation_v05
kznx
0
150
「AI-Ready化」のススメ CDLE LT会 #2
kznx
0
73
MLシステム品質強化のススメ
kznx
0
81
Other Decks in Research
See All in Research
[webinaire du Ciirpo] Agnelages plein air : des éleveurs partagent leurs pratiques
institutdelelevage
PRO
0
210
IA for theory
gpeyre
1
500
JPA2026_NetworkTutorial_JunKashihara
junkashihara
0
230
高性能計算機クラスタを用いた大規模点群処理による森林の単木抽出と構造解析
kentaitakura
1
150
Pretrain Where? Investigating How Pretraining Data Diversity Impacts Geospatial Foundation Model Performance
satai
3
180
LA-Bench 2025:実験指示から実行可能手順を生成するためのデータセット/LA-Bench 2025: A Dataset for Generating Executable Experimental Procedures from Experimental Instructions
stktu
0
200
ICAITD 2026 Oral Presentation
rikkabotan7
0
110
就労支援の現場での生成AI活用
enochan
0
150
CVPR2026論文紹介_VLMにとって良いvision encoderとは何か?Rethinking Model Selection in VLM Through the Lens of Gromov-Wasserstein Distance
kobayashi31
1
270
Être éleveur dans un monde en transition : Comprendre les changements pour mieux les accompagner
institutdelelevage
PRO
0
140
第64回CV・PRML勉強会 論文紹介:Linguistic Priors for Visual Decoupling: Towards Symmetric Vision-Brain Alignment
sokikatayama
0
250
湯村研究室の紹介2026 / yumulab2026
yumulab
0
130
Featured
See All Featured
Building Applications with DynamoDB
mza
96
7.2k
Game over? The fight for quality and originality in the time of robots
wayneb77
1
300
Mind Mapping
helmedeiros
1
400
Building an army of robots
kneath
307
47k
Exploring the relationship between traditional SERPs and Gen AI search
raygrieselhuber
PRO
4
4.4k
BBQ
matthewcrist
90
10k
B2B Lead Gen: Tactics, Traps & Triumph
marketingsoph
1
280
jQuery: Nuts, Bolts and Bling
dougneiner
66
8.6k
Helping Users Find Their Own Way: Creating Modern Search Experiences
danielanewman
31
3.4k
Making the Leap to Tech Lead
cromwellryan
135
10k
Efficient Content Optimization with Google Search Console & Apps Script
katarinadahlin
PRO
1
930
Designing Experiences People Love
moore
143
24k
Transcript
ر؍٦فٓ٦صؚٝח״ 鸐⥋زٓؼحؙך陎ⴽ 痥㔐 .BDIJOF-FBSOJOHNJOVUFT ⚥ꅿㄤ⤥
荈䊹稱➜ ⚥ꅿㄤ⤥ !L[OY • ➬✲ • 尣ꨵ孡䊨噟דاؿز涪噟 • ։ *P5فٓحزؿؓ٦يך涪
• ։ ؽرؔ⠓陽ءأذيך涪 • 㣐㷕ꤍ欰爡⠓➂ぢֽך㣐㷕ꤍ؝٦أ • +"*45 ⻌ꤵ⯓畭猰㷕䪮遭㣐㷕ꤍ㣐㷕 䞔㜠猰㷕ך⽆㡦劍铬玎 • ֿךفٖئٝךⰻ㺁כ⥜㡦灇瑔ك٦أח׃גְתׅ
➙傈׀稱➜ׅⰻ㺁 • ذ٦و • ر؍٦فٓ٦صؚٝד鸐⥋ך*1طحزٙ٦ؙ♳ךزٓؼحؙ陎ⴽ 遤ִךדכזְֲַ • 湡涸 • 넝ְ礵䏝ד*1طحزٙ٦ؙ♳ךزٓؼحؙ陎ⴽ遤ֲص٦ؤ
• ⢽ • *1طحزٙ٦ؙ♳ךزٓؼحؙ陎ⴽ׃ג⮚⯓ⵖ䖴 • 7PJQװؽرؔ⠓陽ך״ֲזأزٔ٦ىؚٝ،فٔ؛٦ءّٝך鸐⥋넝ㅷ颵ד㹋 植 • ⢽ • إُؗٔذ؍錁挿ַ䠐㔳׃זְ،فٔ؛٦ءّٝךزٓؼحؙ黪倖
⡦ָꨇ׃ְךַ • 陎ⴽךꨇ׃ׁ • ه٦ز殢〾זוך䩛ַָ搀׃חزٓؼحؙ窟鎘䞔㜠ֽד陎ⴽ • أزٔ٦ىؚٝ،فٔ؛٦ءّٝכه٦ز殢〾⹛涸ח⢪欽ׅ暴㹀דֹז ְ • ⴽفٗز؝ٕדزٝطؚٔٝ
• 农〾⻉ׁגְ
⟎铡ה胜兝 • ⟎铡 • زٓؼحؙך窟鎘䞔㜠ֽⰅ⸂׃ر؍٦فٓ٦صؚٝך䩛岀⢪ ֲֿהדծزٓؼحؙ陎ⴽך礵䏝ぢ♳דֹךדכזְַ • ر؍٦فٓ٦صؚٝ䩛岀ה׃ג䱰欽׃胜兝 • 鵚䎃ծ堣唒㷕统ך♧䩛岀ד֮ر؍٦فٓ٦صؚٝٝח״陎ⴽ礵䏝
ָぢ♳
*1طحزٙ٦ؙ • *1鸐⥋ • ػ؛حز鸐⥋ • ٕ٦ذ؍ؚٝ穗騟ⵖ 䖴 • ⢽ִהչػ؛حزכ㸓
ꂁ⤑ך♧אך蚚暟꧊ꂁ 䨽װإٝة٦ָ穗騟ⵖ 䖴պ • 5$1 • ؝طؙءّٝ㘗فٗز؝ ٕ • ⱄ鷏ⵖ䖴זו遤ֲ • ⢽ִהչ㼭ⴓֽ׃ג蚚 鸡׃גֻ㸓ꂁ⤑պ *40ꥡ㾴 層 名称 IP Suite 7 アプリケーション層 HTTP, SMTP, DNS 6 プレゼンテーション層 MIME, TLS 5 セッション層 SIP, RTP 4 トランスポート層 TCP, UDP 3 ネットワーク層 IP 2 データリンク層 PPP, L2TP 1 物理層
5$1ؿٗ٦ • 鸐⥋ך㨣ַ穄 ✪תדך♧鸬ךػ ؛حزך鷏「⥋ • ؿٗ٦ך窟鎘䞔㜠 鎘皾דֹ • Ⰻػ؛حز侧
• Ⰻر٦ةꆀ • ؿٗ٦ך竰竲儗 • 䎂㖱ػ؛حزꞿ • 䎂㖱ػ؛حزⵋ滠 ꥫ • ٥٥٥٥٥זו 鷏⥋⩎ 「⥋⩎ 4:/ 4:/"$, "$, '*/"$, "$, '*/"$, "$, ؝طؙءّٝ然甧 ؝طؙءّٝ穄✪ ؝طؙءّٝ㨣 ر٦ة鷏⥋穄✪٥؝طؙءّٝⴖ倖 ؿ ٗ f 儗 穗 麓 %"5" "$, %"5" "$,
陎ⴽך彊⪒ • ر؍٦فٓ٦صؚٝٓ؎ـٓٔ • ,FSBT 5FOTPS'MPX CBDLFOE • ر٦ةإحز •
Ⱅׁגְر٦ةإحزⵃ欽 • /*.4 农〾⻉ׁגְ鸐⥋ろر٦ةإحز • IUUQTQSPKFDUTDTEBMDBQSPKFDUY%PXOMPBEIUNM • صُ٦ٕٓطحزٙ٦ؙ • Ⰻ穠さך㾴ךطحزٙ٦ؙ ؿٗ٦ך窟鎘䞔㜠갪湡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زٓؿ؍حؙ珏겲 5&-/&5 '51 )551 %/4 -JNF 4TI MPDBMGPSXBSEJOH 4TI SFNPUF'PSXBSEJOH 4DQ 4GUQ 9 4IFMM 13*.64 ;'0/&
صُ٦ٕٓ طحزٙ٦ؙ • ؿٗ٦ך窟鎘䞔㜠 Ⰵ⸂ • Ⰻػ؛حز侧 • Ⰻر٦ةꆀ
• ؿٗ٦ך竰竲儗 • 䎂㖱ػ؛حزꞿ • 䎂㖱ػ؛حزⵋ滠 ꥫ • ٥٥٥٥٥זו • زٓؼحؙך珏겲 ⳿⸂ • )551 44) זו
زٖ٦صؚٝ • FQPDI • 7BMVBUJPOMPTT 䖓
穠卓 プロトコル precision recall F1-score support TELNET 1.00 1.00 1.00
97 FTP 1.00 1.00 1.00 83 HTTP 0.99 1.00 0.99 207 DNS 0.97 0.98 0.98 186 lime 0.98 0.96 0.97 210 SSH (local) 1.00 1.00 1.00 212 SSH (remote) 0.99 0.99 0.99 190 Scp 0.98 0.98 0.98 196 Sftp 0.99 0.98 0.98 206 X11 1.00 0.99 1.00 225 Shell 0.98 0.99 0.99 172 PRIMUS 1.00 0.98 0.99 221 ZFONE 0.97 0.99 0.98 195 Avg 0.99 0.99 0.99 2400 • '⦼ • '⦼כQSFDJTJPOהSFDBMM ך锃ㄤ䎂㖱
➭ך䩛岀הך嫰鯰 提案⼿法(DL) C4.5 DR FPR DR FPR TELNET 1.00
0.0 0.998 0.0 FTP 1.00 0.0 0.995 0.0 HTTP 1.00 0.001 0.991 0.0 DNS 0.98 0.003 1.0 0.0 lime 0.96 0.002 0.995 0.001 SSH (local) 1.00 0 1.0 0.0 SSH (remote) 0.99 0.001 1.0 0.0 Scp 0.98 0.002 0.828 0.006 Sftp 0.98 0.001 0.967 0.264 X11 0.99 0 0.714 0.033 Shell 0.99 0.002 0.997 0.0 • ➭ך锷俑ך穠卓ה嫰鯰ׅה嚊 יず瘝 • ♧鿇ךفٗز؝ٕ TDQ Y ד 礵䏝ぢ♳ָ鋅 • 嫰鯰鐰⣣䭷垥 • %3嗚⳿桦3FDBMM • '13铎嗚⳿桦'1 '1 5/ • 嫰鯰锷俑 • "MTIBNNBSJ 3JZBE BOE"/VS ;JODJS)FZXPPE$BO FODSZQUFEUSBGGJDCFJEFOUJGJFE XJUIPVUQPSUOVNCFST *1 BEESFTTFTBOEQBZMPBE JOTQFDUJPO $PNQVUFS OFUXPSLT
➭ך䩛岀הך嫰鯰 #FUUFS #FUUFS
陎ⴽח剣⸬ז窟鎘䞔㜠 • ؿٗ٦Ⰻ⡤ך窟鎘䞔 㜠ָ陎ⴽח剣⸬ • Ⰻػ؛حز侧 • Ⰻر٦ةꆀ • ؿٗ٦ך竰竲儗
• 剑ⴱךꦀ㾴ךⰅ⸂ ךꅾך✳⛦䎂㖱ד 嫰鯰 剣⸬䏝넝
〳鋔⻉׃ג • ⳿⸂㾴湫ךꦀ㾴 侔䋒㔳חفٗحز׃ג • ַזכֹהؙٓ أة٦ⴓꨄדֹגְ 䭁㣐 䭁㣐
礵䏝ぢ♳ • صُ٦ٕٓطحزٙ٦ؙךع؎ػ٦ ػًٓ٦ةثُ٦صؚٝ • ꦀ㾴ךِصحز侧 • "DUJWBUJPO • %SPQPVU
• #BUDI؟؎ؤ • 0QUJNJ[FS • ꅾⴱ劍⻉ • FQPDI侧 • ̔3BOEPNJ[FE4FBSDIד֮אֽ ג穾鴥 パラメータ ベスト Activation relu dropout 0.1 Batch size 500 Optimizer Nadam Init mode Glorot uniform epoch数 1000 • كأزػًٓ٦ة
תה • زٓؼحؙך窟鎘䞔㜠ַֽر؍٦فٓ٦صؚٝך䩛岀 ⢪ג鸐⥋ך*1طحزٙ٦ؙ♳ךزٓؼحؙ陎ⴽ䖞勻䩛 岀״ず瘝⟃♳ך礵䏝ד㹋植דֹ • ع؎ػ٦ػًٓ٦ةثُ٦صؚٝכ礵䏝何㊣ךחכהג 㣐ⴖָ㣐㢌
None