Neural_cryptography_IJCNN09.zip

  • nish1012
    了解作者
  • matlab
    开发工具
  • 192KB
    文件大小
  • zip
    文件格式
  • 0
    收藏次数
  • 1 积分
    下载积分
  • 3
    下载次数
  • 2016-08-03 13:12
    上传日期
artificial neural networks (ANN) is a network inspired by biological neural networks which are used to estimate or approximate functions that can depend on a large number of inputs that are generally unknown.
Neural_cryptography_IJCNN09.zip
  • Neural_cryptography_IJCNN09.pdf
    210.3KB
内容介绍
<html xmlns="http://www.w3.org/1999/xhtml"> <head> <meta charset="utf-8"> <meta name="generator" content="pdf2htmlEX"> <meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1"> <link rel="stylesheet" href="https://static.pudn.com/base/css/base.min.css"> <link rel="stylesheet" href="https://static.pudn.com/base/css/fancy.min.css"> <link rel="stylesheet" href="https://static.pudn.com/prod/directory_preview_static/622b6c7a3d2fbb000785aab5/raw.css"> <script src="https://static.pudn.com/base/js/compatibility.min.js"></script> <script src="https://static.pudn.com/base/js/pdf2htmlEX.min.js"></script> <script> try{ pdf2htmlEX.defaultViewer = new pdf2htmlEX.Viewer({}); }catch(e){} </script> <title></title> </head> <body> <div id="sidebar" style="display: none"> <div id="outline"> </div> </div> <div id="pf1" class="pf w0 h0" data-page-no="1"><div class="pc pc1 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/622b6c7a3d2fbb000785aab5/bg1.jpg"><div class="c x0 y1 w0 h2"><div class="t m0 x1 h3 y2 ff1 fs0 fc0 sc0 ls0 ws0">Proceedings of International Joint Conference on Neural Networks, <span class="_ _0"></span>Atlanta, Georgia, USA, June 14-19, 2009</div><div class="t m0 x1 h3 y3 ff1 fs0 fc0 sc0 ls0 ws0">978-1-4244-3553-1/09/$25.00 &#169;2009 IEEE</div></div><div class="t m0 x2 h4 y4 ff2 fs1 fc1 sc0 ls0 ws0">Impr<span class="_ _0"></span>oved<span class="_ _1"> </span>Security<span class="_ _1"> </span>of<span class="_ _1"> </span>Neural<span class="_ _2"> </span>Cryptograph<span class="_ _3"></span>y<span class="_ _1"> </span>Using</div><div class="t m0 x3 h4 y5 ff2 fs1 fc1 sc0 ls0 ws0">Don&#8217;t-T<span class="_ _4"></span>rust-My-Partner<span class="_ _1"> </span>and<span class="_ _1"> </span>Err<span class="_ _3"></span>or<span class="_ _1"> </span>Prediction</div><div class="t m0 x4 h5 y6 ff3 fs2 fc1 sc0 ls0 ws0">Ahmed<span class="_ _5"> </span>M.<span class="_ _5"> </span>Allam<span class="_ _5"> </span>and<span class="_ _5"> </span>Hazem<span class="_ _5"> </span>M.<span class="_ _5"> </span>Abbas</div><div class="t m0 x5 h6 y7 ff2 fs2 fc1 sc0 ls0 ws0">Mentor<span class="_ _6"> </span>Graphics<span class="_ _6"> </span>Egypt</div><div class="t m0 x6 h5 y8 ff3 fs2 fc1 sc0 ls0 ws0">78<span class="_ _5"> </span>Elnozha<span class="_ _5"> </span>St.,<span class="_ _5"> </span>Helioplois,<span class="_ _5"> </span>Cairo<span class="_ _5"> </span>11341,<span class="_ _5"> </span>Egypt</div><div class="t m0 x7 h7 y9 ff4 fs3 fc1 sc0 ls0 ws0">Abstract<span class="ff2">&#8212;Neural<span class="_ _7"> </span>cryptography<span class="_ _7"> </span>deals<span class="_ _7"> </span>with<span class="_ _8"> </span>the<span class="_ _8"> </span>problem<span class="_ _7"> </span>of</span></div><div class="t m0 x8 h7 ya ff2 fs3 fc1 sc0 ls0 ws0">key<span class="_ _5"> </span>exchange<span class="_ _5"> </span>using<span class="_ _5"> </span>the<span class="_ _9"> </span>mutual<span class="_ _9"> </span>learning<span class="_ _5"> </span>concept<span class="_ _5"> </span>between<span class="_ _5"> </span>two</div><div class="t m0 x8 h7 yb ff2 fs3 fc1 sc0 ls0 ws0">neural<span class="_ _a"> </span>networks.<span class="_ _a"> </span>The<span class="_ _a"> </span>two<span class="_ _a"> </span>netw<span class="_ _3"></span>orks<span class="_ _a"> </span>will<span class="_ _a"> </span>exchange<span class="_ _a"> </span>their<span class="_ _a"> </span>outputs</div><div class="t m0 x8 h7 yc ff2 fs3 fc1 sc0 ls0 ws0">(in<span class="_ _b"> </span>bits)<span class="_ _b"> </span>so<span class="_ _b"> </span>that<span class="_ _b"> </span>the<span class="_ _b"> </span>key<span class="_ _b"> </span>between<span class="_ _b"> </span>the<span class="_ _b"> </span>tw<span class="_ _3"></span>o<span class="_ _b"> </span>communicating</div><div class="t m0 x8 h7 yd ff2 fs3 fc1 sc0 ls0 ws0">parties<span class="_ _6"> </span>is<span class="_ _8"> </span>ev<span class="_ _3"></span>entually<span class="_ _6"> </span>represented<span class="_ _6"> </span>in<span class="_ _8"> </span>the<span class="_ _6"> </span>&#64257;nal<span class="_ _8"> </span>lear<span class="_ _3"></span>ned<span class="_ _6"> </span>weights</div><div class="t m0 x8 h7 ye ff2 fs3 fc1 sc0 ls0 ws0">and<span class="_ _8"> </span>the<span class="_ _8"> </span>two<span class="_ _6"> </span>networks<span class="_ _8"> </span>are<span class="_ _8"> </span>said<span class="_ _8"> </span>to<span class="_ _6"> </span>be<span class="_ _8"> </span>synchronized.<span class="_ _8"> </span>Security</div><div class="t m0 x8 h7 yf ff2 fs3 fc1 sc0 ls0 ws0">of<span class="_ _5"> </span>neural<span class="_ _5"> </span>synchronization<span class="_ _5"> </span>depends<span class="_ _5"> </span>on<span class="_ _6"> </span>the<span class="_ _5"> </span>probability<span class="_ _9"> </span>that<span class="_ _6"> </span>an</div><div class="t m0 x8 h7 y10 ff2 fs3 fc1 sc0 ls0 ws0">attacker<span class="_ _6"> </span>can<span class="_ _6"> </span>synchronize<span class="_ _6"> </span>with<span class="_ _6"> </span>any<span class="_ _6"> </span>of<span class="_ _8"> </span>the<span class="_ _5"> </span>two<span class="_ _6"> </span>parties<span class="_ _6"> </span>during</div><div class="t m0 x8 h7 y11 ff2 fs3 fc1 sc0 ls0 ws0">the<span class="_ _8"> </span>training<span class="_ _8"> </span>pr<span class="_ _3"></span>ocess,<span class="_ _8"> </span>so<span class="_ _8"> </span>decr<span class="_ _0"></span>easing<span class="_ _8"> </span>this<span class="_ _8"> </span>probability<span class="_ _8"> </span>impr<span class="_ _0"></span>oves</div><div class="t m0 x8 h7 y12 ff2 fs3 fc1 sc0 ls0 ws0">the<span class="_ _c"> </span>reliability<span class="_ _a"> </span>of<span class="_ _c"> </span>exchanging<span class="_ _c"> </span>their<span class="_ _c"> </span>output<span class="_ _c"> </span>bits<span class="_ _c"> </span>through<span class="_ _c"> </span>a<span class="_ _c"> </span>public</div><div class="t m0 x8 h7 y13 ff2 fs3 fc1 sc0 ls0 ws0">channel.<span class="_ _5"> </span>This<span class="_ _5"> </span>work<span class="_ _5"> </span>proposes<span class="_ _9"> </span>an<span class="_ _5"> </span>exchange<span class="_ _6"> </span>technique<span class="_ _5"> </span>that<span class="_ _5"> </span>will</div><div class="t m0 x8 h7 y14 ff2 fs3 fc1 sc0 ls0 ws0">disrupt<span class="_"> </span>the<span class="_"> </span>attacker<span class="_"> </span>con&#64257;dence<span class="_ _d"> </span>in<span class="_"> </span>the<span class="_"> </span>exchanged<span class="_ _d"> </span>outputs<span class="_"> </span>during</div><div class="t m0 x8 h7 y15 ff2 fs3 fc1 sc0 ls0 ws0">training.<span class="_"> </span>The<span class="_"> </span>algorithm<span class="_"> </span>is<span class="_"> </span>based<span class="_"> </span>on<span class="_"> </span>one<span class="_"> </span>party<span class="_"> </span>sending<span class="_"> </span>err<span class="_ _0"></span>oneous</div><div class="t m0 x8 h7 y16 ff2 fs3 fc1 sc0 ls0 ws0">output<span class="_ _6"> </span>bits<span class="_ _8"> </span>with<span class="_ _6"> </span>the<span class="_ _6"> </span>other<span class="_ _8"> </span>party<span class="_ _6"> </span>being<span class="_ _8"> </span>capable<span class="_ _6"> </span>of<span class="_ _6"> </span>predicting</div><div class="t m0 x8 h7 y17 ff2 fs3 fc1 sc0 ls0 ws0">and<span class="_ _5"> </span>removing<span class="_ _9"> </span>this<span class="_ _5"> </span>error<span class="_ _e"></span>.<span class="_ _5"> </span>The<span class="_ _5"> </span>proposed<span class="_ _5"> </span>approach<span class="_ _9"> </span>is<span class="_ _5"> </span>shown<span class="_ _5"> </span>to</div><div class="t m0 x8 h7 y18 ff2 fs3 fc1 sc0 ls0 ws0">outperform<span class="_ _d"> </span>the<span class="_ _a"> </span>synchronization<span class="_ _d"> </span>with<span class="_ _a"> </span>feedback<span class="_ _a"> </span>algorithm<span class="_ _a"> </span>in<span class="_ _a"> </span>the</div><div class="t m0 x8 h7 y19 ff2 fs3 fc1 sc0 ls0 ws0">time<span class="_ _c"> </span>needed<span class="_ _9"> </span>for<span class="_ _c"> </span>the<span class="_ _c"> </span>parties<span class="_ _9"> </span>to<span class="_ _c"> </span>synchronize.</div><div class="t m0 x9 h8 y1a ff3 fs4 fc1 sc0 ls0 ws0">I<span class="_ _f"></span>.<span class="_ _10"> </span>I<span class="_ _f"></span><span class="fs5">N<span class="_ _f"></span>T<span class="_ _f"></span>RO<span class="_ _f"></span>D<span class="_ _f"></span>U<span class="_ _f"></span>C<span class="_ _f"></span>T<span class="_ _f"></span>I<span class="_ _f"></span>O<span class="_ _f"></span>N</span></div><div class="t m0 x7 h8 y1b ff3 fs4 fc1 sc0 ls0 ws0">Neural<span class="_ _8"> </span>networks<span class="_ _8"> </span>(NNs)<span class="_ _8"> </span>are<span class="_ _8"> </span>able<span class="_ _7"> </span>to<span class="_ _8"> </span>solve<span class="_ _8"> </span>so<span class="_ _8"> </span>called<span class="_ _8"> </span>non</div><div class="t m0 x8 h8 y1c ff3 fs4 fc1 sc0 ls0 ws0">formalized<span class="_ _10"> </span>problems<span class="_ _10"> </span>or<span class="_ _10"> </span>weakly<span class="_ _10"> </span>formalized<span class="_ _10"> </span>problems<span class="_ _10"> </span>that</div><div class="t m0 x8 h8 y1d ff3 fs4 fc1 sc0 ls0 ws0">requires<span class="_ _9"> </span>learning<span class="_ _5"> </span>process<span class="_ _5"> </span>based<span class="_ _5"> </span>on<span class="_ _5"> </span>a<span class="_ _9"> </span>real<span class="_ _5"> </span>experimental<span class="_ _9"> </span>data</div><div class="t m0 x8 h8 y1e ff3 fs4 fc1 sc0 ls0 ws0">[1].<span class="_"> </span>Supervised<span class="_ _a"> </span>NNs<span class="_ _a"> </span>models<span class="_ _a"> </span>are<span class="_ _a"> </span>trained<span class="_ _a"> </span>on<span class="_ _a"> </span>input/output<span class="_ _a"> </span>pairs</div><div class="t m0 x8 h8 y1f ff3 fs4 fc1 sc0 ls0 ws0">to<span class="_ _c"> </span>achiev<span class="_ _3"></span>e<span class="_ _c"> </span>a<span class="_ _9"> </span>certain<span class="_ _c"> </span>task.<span class="_ _c"> </span>This<span class="_ _9"> </span>training<span class="_ _c"> </span>is<span class="_ _c"> </span>based<span class="_ _9"> </span>on<span class="_ _c"> </span>adjusting</div><div class="t m0 x8 h8 y20 ff3 fs4 fc1 sc0 ls0 ws0">the<span class="_ _6"> </span>initial<span class="_ _6"> </span>randomized<span class="_ _6"> </span>synaptic<span class="_ _6"> </span>weights<span class="_ _6"> </span>by<span class="_ _8"> </span>applying<span class="_ _5"> </span>a<span class="_ _6"> </span>pre-</div><div class="t m0 x8 h8 y21 ff3 fs4 fc1 sc0 ls0 ws0">de&#64257;ned<span class="_ _8"> </span>learning<span class="_ _6"> </span>rule.<span class="_ _8"> </span>T<span class="_ _e"></span>wo<span class="_ _6"> </span>NNs<span class="_ _8"> </span>having<span class="_ _6"> </span>the<span class="_ _8"> </span>same<span class="_ _6"> </span>structure</div><div class="t m0 x8 h8 y22 ff3 fs4 fc1 sc0 ls0 ws0">and<span class="_ _6"> </span>dif<span class="_ _3"></span>ferent<span class="_ _5"> </span>initial<span class="_ _6"> </span>synaptic<span class="_ _6"> </span>weights<span class="_ _6"> </span>can<span class="_ _6"> </span>do<span class="_ _6"> </span>the<span class="_ _6"> </span>same<span class="_ _6"> </span>task</div><div class="t m0 x8 h8 y23 ff3 fs4 fc1 sc0 ls0 ws0">if<span class="_ _9"> </span>both<span class="_ _9"> </span>are<span class="_ _5"> </span>trained<span class="_ _9"> </span>on<span class="_ _9"> </span>the<span class="_ _5"> </span>same<span class="_ _9"> </span>input/output<span class="_ _5"> </span>pairs<span class="_ _9"> </span>while<span class="_ _9"> </span>the</div><div class="t m0 x8 h8 y24 ff3 fs4 fc1 sc0 ls0 ws0">&#64257;nal<span class="_ _5"> </span>synaptic<span class="_ _5"> </span>weights<span class="_ _5"> </span>of<span class="_ _5"> </span>the<span class="_ _5"> </span>two<span class="_ _5"> </span>networks<span class="_ _9"> </span>need<span class="_ _5"> </span>not<span class="_ _5"> </span>be<span class="_ _5"> </span>the</div><div class="t m0 x8 h8 y25 ff3 fs4 fc1 sc0 ls0 ws0">same.<span class="_ _6"> </span>In<span class="_ _8"> </span>fact<span class="_ _6"> </span>this<span class="_ _6"> </span>phenomenon<span class="_ _8"> </span>is<span class="_ _6"> </span>very<span class="_ _6"> </span>interesting<span class="_ _8"> </span>and<span class="_ _6"> </span>can</div><div class="t m0 x8 h8 y26 ff3 fs4 fc1 sc0 ls0 ws0">be<span class="_ _5"> </span>modi&#64257;ed<span class="_ _5"> </span>to<span class="_ _5"> </span>achieve<span class="_ _9"> </span>another<span class="_ _5"> </span>goal,<span class="_ _5"> </span>i.e.,<span class="_ _6"> </span>the<span class="_ _5"> </span>two<span class="_ _5"> </span>networks</div><div class="t m0 x8 h8 y27 ff3 fs4 fc1 sc0 ls0 ws0">hav<span class="_ _0"></span>e<span class="_ _6"> </span>the<span class="_ _5"> </span>same<span class="_ _5"> </span>&#64257;nal<span class="_ _5"> </span>weights.<span class="_ _5"> </span>One<span class="_ _6"> </span>way<span class="_ _5"> </span>to<span class="_ _5"> </span>do<span class="_ _5"> </span>that<span class="_ _5"> </span>is<span class="_ _5"> </span>for<span class="_ _6"> </span>the</div><div class="t m0 x8 h8 y28 ff3 fs4 fc1 sc0 ls0 ws0">two<span class="_ _6"> </span>networks<span class="_ _6"> </span>to<span class="_ _6"> </span>be<span class="_ _8"> </span>presented<span class="_ _6"> </span>with<span class="_ _6"> </span>common<span class="_ _8"> </span>input<span class="_ _6"> </span>patterns</div><div class="t m0 x8 h8 y29 ff3 fs4 fc1 sc0 ls0 ws0">while<span class="_ _6"> </span>being<span class="_ _6"> </span>trained<span class="_ _8"> </span>on<span class="_ _5"> </span>the<span class="_ _8"> </span>output<span class="_ _5"> </span>of<span class="_ _8"> </span>each<span class="_ _5"> </span>other<span class="_ _8"> </span>instead<span class="_ _6"> </span>of</div><div class="t m0 x8 h8 y2a ff3 fs4 fc1 sc0 ls0 ws0">prede&#64257;ned<span class="_ _8"> </span>tar<span class="_ _0"></span>get<span class="_ _8"> </span>patterns.<span class="_ _8"> </span>The<span class="_ _8"> </span>applied<span class="_ _6"> </span>learning<span class="_ _8"> </span>rule<span class="_ _8"> </span>needs</div><div class="t m0 x8 h8 y2b ff3 fs4 fc1 sc0 ls0 ws0">to<span class="_ _c"> </span>be<span class="_ _c"> </span>so<span class="_ _c"> </span>ef&#64257;cient<span class="_ _c"> </span>that<span class="_ _c"> </span>the<span class="_ _c"> </span>two<span class="_ _c"> </span>synaptic<span class="_ _9"> </span>weight<span class="_ _c"> </span>vectors<span class="_ _c"> </span>of<span class="_ _c"> </span>the</div><div class="t m0 x8 h8 y2c ff3 fs4 fc1 sc0 ls0 ws0">two<span class="_"> </span>networks<span class="_"> </span>become<span class="_"> </span>close<span class="_"> </span>to<span class="_"> </span>each<span class="_"> </span>other<span class="_"> </span>and<span class="_"> </span>thus<span class="_"> </span>correlated.</div><div class="t m0 x8 h8 y2d ff3 fs4 fc1 sc0 ls0 ws0">Hence,<span class="_ _a"> </span>the<span class="_ _a"> </span>&#64257;nal<span class="_ _c"> </span>two<span class="_"> </span>weight<span class="_ _c"> </span>vectors<span class="_"> </span>are<span class="_ _c"> </span>almost<span class="_ _a"> </span>identical.<span class="_ _a"> </span>The</div><div class="t m0 x8 h8 y2e ff3 fs4 fc1 sc0 ls0 ws0">correlation<span class="_ _7"> </span>between<span class="_ _7"> </span>the<span class="_ _11"> </span>two<span class="_ _7"> </span>weight<span class="_ _7"> </span>vectors<span class="_ _7"> </span>is<span class="_ _7"> </span>also<span class="_ _11"> </span>called</div><div class="t m0 x8 h8 y2f ff3 fs4 fc1 sc0 ls0 ws0">the<span class="_ _9"> </span>ov<span class="_ _3"></span>erlap.<span class="_ _9"> </span>When<span class="_ _9"> </span>the<span class="_ _9"> </span>overlap<span class="_ _9"> </span>is<span class="_ _9"> </span><span class="ff5">100%<span class="_ _9"> </span></span>(i.e.<span class="_ _9"> </span>the<span class="_ _9"> </span>two<span class="_ _9"> </span>weight</div><div class="t m0 x8 h8 y30 ff3 fs4 fc1 sc0 ls0 ws0">vectors<span class="_ _6"> </span>are<span class="_ _8"> </span>identical)<span class="_ _6"> </span>it<span class="_ _8"> </span>can<span class="_ _6"> </span>be<span class="_ _8"> </span>said<span class="_ _6"> </span>that<span class="_ _8"> </span>the<span class="_ _6"> </span>two<span class="_ _8"> </span>networks</div><div class="t m0 x8 h8 y31 ff3 fs4 fc1 sc0 ls0 ws0">hav<span class="_ _0"></span>e<span class="_ _9"> </span>synchronized<span class="_ _9"> </span>with<span class="_ _9"> </span>each<span class="_ _9"> </span>other<span class="_ _0"></span>.</div><div class="t m0 x7 h8 y32 ff3 fs4 fc1 sc0 ls0 ws0">An<span class="_ _b"> </span>aim<span class="_ _b"> </span>of<span class="_ _b"> </span>cryptography<span class="_ _b"> </span>is<span class="_ _b"> </span>to<span class="_ _b"> </span>transmit<span class="_ _b"> </span>a<span class="_ _b"> </span>secret<span class="_ _b"> </span>mes-</div><div class="t m0 x8 h8 y33 ff3 fs4 fc1 sc0 ls0 ws0">sage<span class="_ _10"> </span>between<span class="_ _10"> </span>two<span class="_ _11"> </span>partners,<span class="_ _10"> </span>A<span class="_ _10"> </span>and<span class="_ _10"> </span>B,<span class="_ _10"> </span>while<span class="_ _10"> </span>an<span class="_ _10"> </span>attacker<span class="_ _0"></span>,</div><div class="t m0 x8 h8 y34 ff3 fs4 fc1 sc0 ls0 ws0">E,<span class="_ _6"> </span>who<span class="_ _6"> </span>happens<span class="_ _6"> </span>to<span class="_ _6"> </span>access<span class="_ _6"> </span>the<span class="_ _6"> </span>communication<span class="_ _6"> </span>channel<span class="_ _6"> </span>will</div><div class="t m0 x8 h8 y35 ff3 fs4 fc1 sc0 ls0 ws0">not<span class="_ _7"> </span>be<span class="_ _11"> </span>able<span class="_ _7"> </span>to<span class="_ _11"> </span>&#64257;gure<span class="_ _7"> </span>out<span class="_ _11"> </span>the<span class="_ _11"> </span>context<span class="_ _8"> </span>of<span class="_ _11"> </span>this<span class="_ _7"> </span>message.<span class="_ _11"> </span>A</div><div class="t m0 x8 h8 y36 ff3 fs4 fc1 sc0 ls0 ws0">number<span class="_ _11"> </span>of<span class="_ _7"> </span>methods<span class="_ _11"> </span>hav<span class="_ _3"></span>e<span class="_ _7"> </span>been<span class="_ _11"> </span>introduced<span class="_ _11"> </span>to<span class="_ _11"> </span>achie<span class="_ _0"></span>ve<span class="_ _11"> </span>this</div><div class="t m0 x8 h8 y37 ff3 fs4 fc1 sc0 ls0 ws0">goal<span class="_ _10"> </span>[2][3][4][5].<span class="_ _10"> </span>In<span class="_ _10"> </span>1976<span class="_ _10"> </span>Dif&#64257;e<span class="_ _10"> </span>and<span class="_ _10"> </span>Hellman<span class="_ _10"> </span>dev<span class="_ _0"></span>eloped</div><div class="t m0 x8 h8 y38 ff3 fs4 fc1 sc0 ls0 ws0">a<span class="_ _7"> </span>mechanism<span class="_ _11"> </span>based<span class="_ _7"> </span>on<span class="_ _11"> </span>number<span class="_ _7"> </span>theory<span class="_ _7"> </span>by<span class="_ _11"> </span>which<span class="_ _7"> </span>a<span class="_ _11"> </span>secret</div><div class="t m0 xa h8 y9 ff3 fs4 fc1 sc0 ls0 ws0">ke<span class="_ _3"></span>y<span class="_ _9"> </span>can<span class="_ _9"> </span>be<span class="_ _9"> </span>exchanged<span class="_ _9"> </span>by<span class="_ _5"> </span>two<span class="_ _9"> </span>parties<span class="_ _9"> </span>ov<span class="_ _3"></span>er<span class="_ _9"> </span>a<span class="_ _9"> </span>public<span class="_ _9"> </span>channel</div><div class="t m0 xa h8 y39 ff3 fs4 fc1 sc0 ls0 ws0">which<span class="_ _c"> </span>is<span class="_ _c"> </span>accessible<span class="_ _9"> </span>to<span class="_ _c"> </span>any<span class="_ _c"> </span>attacker<span class="_ _c"> </span>[2][3].<span class="_ _c"> </span>Alternati<span class="_ _0"></span>vely<span class="_ _0"></span>,<span class="_ _c"> </span>two</div><div class="t m0 xa h8 y3a ff3 fs4 fc1 sc0 ls0 ws0">networks<span class="_ _7"> </span>trained<span class="_ _7"> </span>on<span class="_ _11"> </span>their<span class="_ _7"> </span>outputs<span class="_ _11"> </span>are<span class="_ _7"> </span>able<span class="_ _7"> </span>to<span class="_ _11"> </span>achiev<span class="_ _0"></span>e<span class="_ _11"> </span>the</div><div class="t m0 xa h8 y3b ff3 fs4 fc1 sc0 ls0 ws0">same<span class="_ _5"> </span>objectiv<span class="_ _3"></span>e<span class="_ _5"> </span>by<span class="_ _5"> </span>means<span class="_ _6"> </span>of<span class="_ _5"> </span>mutual<span class="_ _6"> </span>learning<span class="_ _5"> </span>[6].<span class="_ _6"> </span>The<span class="_ _5"> </span>most</div><div class="t m0 xa h8 y3c ff3 fs4 fc1 sc0 ls0 ws0">common<span class="_ _c"> </span>model<span class="_ _a"> </span>used<span class="_ _c"> </span>in<span class="_ _c"> </span>neural<span class="_ _c"> </span>cryptography<span class="_ _a"> </span>is<span class="_ _c"> </span>known<span class="_ _a"> </span>as<span class="_ _c"> </span>the</div><div class="t m0 xa h8 yf ff3 fs4 fc1 sc0 ls0 ws0">T<span class="_ _0"></span>ree<span class="_"> </span>parity<span class="_ _d"> </span>Machine<span class="_ _a"> </span>(TPM)<span class="_"> </span>since<span class="_"> </span>it<span class="_"> </span>keeps<span class="_"> </span>the<span class="_"> </span>state<span class="_"> </span>of<span class="_"> </span>the<span class="_"> </span>two</div><div class="t m0 xa h8 y3d ff3 fs4 fc1 sc0 ls0 ws0">parties<span class="_ _5"> </span>secret,<span class="_ _5"> </span>and<span class="_ _5"> </span>thus<span class="_ _5"> </span>it<span class="_ _5"> </span>is<span class="_ _5"> </span>more<span class="_ _5"> </span>secure<span class="_ _5"> </span>than<span class="_ _5"> </span>using<span class="_ _5"> </span>simple</div><div class="t m0 xa h8 y3e ff3 fs4 fc1 sc0 ls0 ws0">network.</div><div class="t m0 xb h8 y3f ff3 fs4 fc1 sc0 ls0 ws0">The<span class="_ _11"> </span>aim<span class="_ _11"> </span>of<span class="_ _11"> </span>this<span class="_ _11"> </span>work<span class="_ _11"> </span>is<span class="_ _11"> </span>to<span class="_ _11"> </span>introduce<span class="_ _11"> </span>a<span class="_ _11"> </span>mechanism<span class="_ _11"> </span>to</div><div class="t m0 xa h8 y40 ff3 fs4 fc1 sc0 ls0 ws0">improv<span class="_ _0"></span>e<span class="_ _b"> </span>the<span class="_ _b"> </span>security<span class="_ _b"> </span>of<span class="_ _b"> </span>the<span class="_ _b"> </span>mutual<span class="_ _b"> </span>learning<span class="_ _b"> </span>process,<span class="_ _b"> </span>so</div><div class="t m0 xa h8 y41 ff3 fs4 fc1 sc0 ls0 ws0">that<span class="_ _11"> </span>the<span class="_ _11"> </span>attacker<span class="_ _11"> </span>&#64257;nd<span class="_ _11"> </span>it<span class="_ _7"> </span>more<span class="_ _11"> </span>dif&#64257;cult<span class="_ _11"> </span>in<span class="_ _11"> </span>listening<span class="_ _11"> </span>to<span class="_ _11"> </span>the</div><div class="t m0 xa h8 y42 ff3 fs4 fc1 sc0 ls0 ws0">communication<span class="_ _8"> </span>between<span class="_ _7"> </span>the<span class="_ _8"> </span>two<span class="_ _8"> </span>parties<span class="_ _8"> </span>during<span class="_ _7"> </span>the<span class="_ _8"> </span>period</div><div class="t m0 xa h8 y43 ff3 fs4 fc1 sc0 ls0 ws0">in<span class="_ _9"> </span>which<span class="_ _9"> </span>the<span class="_ _3"></span>y<span class="_ _9"> </span>increase<span class="_ _c"> </span>their<span class="_ _9"> </span>weight<span class="_ _9"> </span>vectors<span class="_ _9"> </span>ov<span class="_ _0"></span>erlap.</div><div class="t m0 xb h8 y44 ff3 fs4 fc1 sc0 ls0 ws0">The<span class="_ _9"> </span>paper<span class="_ _9"> </span>is<span class="_ _5"> </span>organized<span class="_ _9"> </span>as<span class="_ _9"> </span>follows.<span class="_ _9"> </span>Section<span class="_ _9"> </span>II<span class="_ _9"> </span>presents<span class="_ _9"> </span>an</div><div class="t m0 xa h8 y45 ff3 fs4 fc1 sc0 ls0 ws0">introduction<span class="_ _a"> </span>to<span class="_ _a"> </span>mutual<span class="_ _a"> </span>learning<span class="_ _a"> </span>in<span class="_ _a"> </span>both<span class="_ _a"> </span>a<span class="_ _a"> </span>simple<span class="_ _a"> </span>network<span class="_ _a"> </span>and</div><div class="t m0 xa h8 y46 ff3 fs4 fc1 sc0 ls0 ws0">TPM.<span class="_ _6"> </span>Section<span class="_ _8"> </span>III<span class="_ _6"> </span>sho<span class="_ _3"></span>ws<span class="_ _6"> </span>a<span class="_ _8"> </span>summary<span class="_ _6"> </span>to<span class="_ _6"> </span>most<span class="_ _8"> </span>kno<span class="_ _0"></span>wn<span class="_ _8"> </span>attack</div><div class="t m0 xa h8 y47 ff3 fs4 fc1 sc0 ls0 ws0">against<span class="_ _6"> </span>mutual<span class="_ _8"> </span>learning.<span class="_ _8"> </span>In<span class="_ _6"> </span>section<span class="_ _8"> </span>IV,<span class="_ _6"> </span>a<span class="_ _8"> </span>brief<span class="_ _8"> </span>e<span class="_ _0"></span>xplanation</div><div class="t m0 xa h8 y48 ff3 fs4 fc1 sc0 ls0 ws0">for<span class="_ _8"> </span>neural<span class="_ _8"> </span>synchronization<span class="_ _8"> </span>with<span class="_ _7"> </span>feedback<span class="_ _8"> </span>[7]<span class="_ _8"> </span>is<span class="_ _8"> </span>presented.</div><div class="t m0 xa h8 y49 ff3 fs4 fc1 sc0 ls0 ws0">This<span class="_ _6"> </span>method<span class="_ _5"> </span>was<span class="_ _6"> </span>dev<span class="_ _0"></span>eloped<span class="_ _6"> </span>to<span class="_ _6"> </span>improve<span class="_ _5"> </span>the<span class="_ _6"> </span>security<span class="_ _6"> </span>of<span class="_ _6"> </span>the</div><div class="t m0 xa h8 y4a ff3 fs4 fc1 sc0 ls0 ws0">mutual<span class="_ _6"> </span>learning<span class="_ _6"> </span>process<span class="_ _6"> </span>for<span class="_ _6"> </span>the<span class="_ _6"> </span>TPM<span class="_ _6"> </span>model.<span class="_ _6"> </span>In<span class="_ _6"> </span>section<span class="_ _6"> </span>V,</div><div class="t m0 xa h8 y4b ff3 fs4 fc1 sc0 ls0 ws0">the<span class="_ _8"> </span>DTMP<span class="_ _6"> </span>(Don&#8217;t<span class="_ _6"> </span>Trust<span class="_ _6"> </span>My<span class="_ _8"> </span>P<span class="_ _0"></span>artner)<span class="_ _8"> </span>with<span class="_ _8"> </span>error<span class="_ _6"> </span>prediction</div><div class="t m0 xa h8 y4c ff3 fs4 fc1 sc0 ls0 ws0">approach<span class="_ _9"> </span>is<span class="_ _c"> </span>proposed<span class="_ _9"> </span>to<span class="_ _c"> </span>improve<span class="_ _c"> </span>the<span class="_ _9"> </span>security<span class="_ _c"> </span>of<span class="_ _9"> </span>exchanging</div><div class="t m0 xa h8 y4d ff3 fs4 fc1 sc0 ls0 ws0">the<span class="_ _9"> </span>two<span class="_ _c"> </span>parties<span class="_ _9"> </span>output<span class="_ _9"> </span>bits.<span class="_ _9"> </span>Section<span class="_ _9"> </span>VI<span class="_ _9"> </span>presents<span class="_ _c"> </span>the<span class="_ _9"> </span>possible</div><div class="t m0 xa h8 y4e ff3 fs4 fc1 sc0 ls0 ws0">break-on<span class="_ _c"> </span>scenarios<span class="_ _a"> </span>against<span class="_ _c"> </span>the<span class="_ _c"> </span>proposed<span class="_ _a"> </span>method.<span class="_ _c"> </span>Finally<span class="_ _0"></span>,<span class="_ _c"> </span>the</div><div class="t m0 xa h8 y4f ff3 fs4 fc1 sc0 ls0 ws0">experimental<span class="_ _c"> </span>results<span class="_ _9"> </span>are<span class="_ _9"> </span>shown<span class="_ _c"> </span>in<span class="_ _9"> </span>section<span class="_ _9"> </span>VII.</div><div class="t m0 xc h8 y50 ff3 fs4 fc1 sc0 ls0 ws0">I<span class="_ _f"></span>I<span class="_ _f"></span>.<span class="_ _10"> </span>M<span class="_ _f"></span><span class="fs5">U<span class="_ _f"></span>T<span class="_ _f"></span>UA<span class="_ _f"></span>L<span class="_ _c"> </span>L<span class="_ _f"></span>E<span class="_ _f"></span>A<span class="_ _f"></span>R<span class="_ _f"></span>N<span class="_ _f"></span>I<span class="_ _f"></span>N<span class="_ _f"></span>G<span class="_ _a"> </span>I<span class="_ _f"></span>N<span class="_ _a"> </span></span>T<span class="_ _f"></span><span class="fs5">R<span class="_ _f"></span>E<span class="_ _f"></span>E<span class="_ _c"> </span></span>P<span class="_ _3"></span><span class="fs5">A<span class="_ _f"></span>R<span class="_ _f"></span>I<span class="_ _f"></span>T<span class="_ _f"></span>Y<span class="_ _a"> </span><span class="fs4">M<span class="_ _f"></span></span>A<span class="_ _f"></span>C<span class="_ _f"></span>H<span class="_ _f"></span>I<span class="_ _f"></span>N<span class="_ _f"></span>E<span class="_ _f"></span>S</span></div><div class="t m0 xd h8 y51 ff3 fs4 fc1 sc0 ls0 ws0">(<span class="_ _f"></span>T<span class="_ _f"></span>P<span class="_ _f"></span>M<span class="_ _f"></span>)</div><div class="t m0 xb h8 y52 ff3 fs4 fc1 sc0 ls0 ws0">The<span class="_ _b"> </span>TPM<span class="_ _b"> </span>is<span class="_ _b"> </span>an<span class="_ _b"> </span>ef<span class="_ _3"></span>&#64257;cient<span class="_ _b"> </span>network<span class="_ _b"> </span>structure<span class="_ _10"> </span>employing</div><div class="t m0 xa h8 y53 ff3 fs4 fc1 sc0 ls0 ws0">the<span class="_ _7"> </span>mutual<span class="_ _11"> </span>learning<span class="_ _7"> </span>concept<span class="_ _7"> </span>for<span class="_ _11"> </span>neural<span class="_ _7"> </span>cryptography<span class="_ _0"></span>.<span class="_ _7"> </span>The</div><div class="t m0 xa h8 y54 ff3 fs4 fc1 sc0 ls0 ws0">method<span class="_ _a"> </span>of<span class="_ _a"> </span>mutual<span class="_ _a"> </span>learning<span class="_ _a"> </span>is<span class="_ _a"> </span>explained<span class="_"> </span>below<span class="_"> </span>using<span class="_ _a"> </span>a<span class="_ _a"> </span>simple</div><div class="t m0 xa h8 y55 ff3 fs4 fc1 sc0 ls0 ws0">perceptron<span class="_ _9"> </span>network<span class="_ _c"> </span>structure.</div><div class="t m0 xa h8 y56 ff3 fs4 fc1 sc0 ls0 ws0">Figure1<span class="_ _5"> </span>shows<span class="_ _5"> </span>two<span class="_ _6"> </span>perceptrons<span class="_ _5"> </span>receiving<span class="_ _5"> </span>the<span class="_ _6"> </span>same<span class="_ _6"> </span>random</div><div class="t m0 xa h8 y57 ff3 fs4 fc1 sc0 ls0 ws0">input<span class="_ _8"> </span>pattern<span class="_ _7"> </span><span class="ff6">x</span>.<span class="_ _8"> </span>Their<span class="_ _7"> </span>weights,<span class="_ _8"> </span><span class="ff6">w<span class="_ _f"></span></span>,<span class="_ _8"> </span>are<span class="_ _8"> </span>adjusted<span class="_ _7"> </span>according</div><div class="t m0 xa h8 y58 ff3 fs4 fc1 sc0 ls0 ws0">to<span class="_ _7"> </span>their<span class="_ _11"> </span>mutual<span class="_ _7"> </span>output<span class="_ _11"> </span>bits.<span class="_ _7"> </span>The<span class="_ _11"> </span>output<span class="_ _7"> </span>bit<span class="_ _11"> </span><span class="ff7">&#963;<span class="_ _11"> </span></span>of<span class="_ _11"> </span>a<span class="_ _7"> </span>single</div><div class="t m0 xa h8 y59 ff3 fs4 fc1 sc0 ls0 ws0">perceptron<span class="_ _9"> </span>is<span class="_ _9"> </span>gi<span class="_ _0"></span>ven<span class="_ _9"> </span>by</div><div class="t m0 xe h9 y5a ff7 fs4 fc1 sc0 ls0 ws0">&#963;<span class="_ _c"> </span><span class="ff5">=<span class="_ _d"> </span></span>sig<span class="_ _f"></span>n<span class="ff5">(<span class="ff6">w</span></span></div><div class="t m0 xf ha y5b ff8 fs6 fc1 sc0 ls0 ws0">T</div><div class="t m0 x10 h8 y5a ff7 fs4 fc1 sc0 ls0 ws0">.<span class="ff6">x<span class="ff5">)<span class="_ _12"> </span><span class="ff3">(1)</span></span></span></div><div class="t m0 xa h8 y5c ff3 fs4 fc1 sc0 ls0 ws0">where<span class="_ _9"> </span><span class="ff7">sig<span class="_ _f"></span>n<span class="ff5">(</span>n<span class="ff5">)<span class="_ _d"> </span>=</span></span></div><div class="t m0 x11 hb y5d ff9 fs4 fc1 sc0 ls0 ws0">&#58906;</div><div class="t m0 x12 h8 y5e ff5 fs4 fc1 sc0 ls0 ws0">1<span class="ff7">,<span class="_ _13"> </span><span class="ff3">for<span class="_ _a"> </span></span>n<span class="_ _a"> </span><span class="ffa">&#8805;<span class="_ _a"> </span></span></span>0;</div><div class="t m0 x12 h8 y5f ffa fs4 fc1 sc0 ls0 ws0">&#8722;<span class="ff5">1<span class="ff7">,<span class="_ _14"> </span><span class="ff3">otherwise</span></span></span></div><div class="t m0 xb h8 y34 ff3 fs4 fc1 sc0 ls0 ws0">The<span class="_ _10"> </span>input<span class="_ _b"> </span>pattern,<span class="_ _10"> </span><span class="ff6">x</span>,<span class="_ _b"> </span>is<span class="_ _10"> </span>an<span class="_ _b"> </span><span class="ff7">N<span class="_ _15"></span></span>-component<span class="_ _10"> </span>vector<span class="_ _b"> </span>with</div><div class="t m0 xa h8 y35 ff3 fs4 fc1 sc0 ls0 ws0">its<span class="_ _11"> </span>component<span class="_ _7"> </span>being<span class="_ _11"> </span>restricted<span class="_ _11"> </span>to<span class="_ _11"> </span>be<span class="_ _11"> </span>drawn<span class="_ _7"> </span>from<span class="_ _11"> </span>a<span class="_ _11"> </span>zero-</div><div class="t m0 xa h8 y36 ff3 fs4 fc1 sc0 ls0 ws0">mean<span class="_ _c"> </span>unit-variance<span class="_ _c"> </span>Gaussian<span class="_ _c"> </span>distrib<span class="_ _3"></span>ution.<span class="_ _c"> </span>The<span class="_ _c"> </span>weight<span class="_ _c"> </span>vector</div><div class="t m0 xa h8 y37 ff6 fs4 fc1 sc0 ls0 ws0">w<span class="_ _7"> </span><span class="ff3">is<span class="_ _11"> </span>an<span class="_ _7"> </span>N-dimension<span class="_ _7"> </span>vector<span class="_ _7"> </span>with<span class="_ _7"> </span>continuous<span class="_ _7"> </span>components</span></div><div class="t m0 xa h8 y38 ff3 fs4 fc1 sc0 ls0 ws0">which<span class="_ _c"> </span>are<span class="_ _a"> </span>restricted<span class="_ _c"> </span>to<span class="_ _a"> </span>be<span class="_ _c"> </span>normalized,<span class="_ _c"> </span>since<span class="_ _c"> </span>only<span class="_ _a"> </span>normalized</div><div class="t m0 x13 hc y60 ffb fs7 fc2 sc0 ls0 ws0">121</div></div><div class="pi" data-data='{"ctm":[1.568627,0.000000,0.000000,1.568627,0.000000,0.000000]}'></div></div> </body> </html>
评论
    相关推荐
    • ANN.rar
      This toolbox contains C++ files that implement the Approximate Nearest Neighbor Library (ANN) and and its utility functions in matlab.
    • ANN.rar
      It is a MATLAB code for artificial neural network
    • ANN_TB.zip
      matlabproject is uploading in to pudn
    • ANN.rar
      希望输入不同位置下的T时网络输出为1,输入L时为0,可选网络结构为9-3-1.用BP算法求出权系数和阈值.
    • ANN.zip
      神经网络入门资料,简单易懂,基础知识,入门
    • ANN
      人工神经网络
    • 309780805ANN
      国外经典神经网络设计,神经网络书本附带的Matlab源码,与大家分享!
    • Neural ANN IM.rar
      The identification model of the adapt network was created by Simulink and S-function. The algorithm is a simple and intuitionalist simulation result shows that the speed have better performance
    • ANN.rar
      机器学习,ann python 实现,不依赖于api实现
    • ANN.rar
      ANN算法是一个用C ++编程语言编写的库,以支持两者在各种空间中搜索的精确和近似最近邻居尺寸。它由大学的David M. Mount实施马里兰州和香港科技大学的Sunil Arya技术。 ANN(发音如名为“Ann”)代表近似的最近邻居...