Кластеризация групп входящих пакетов с помощью нейронных сетей конкурирующего типа

InputValues : Array of Extended;

OutputValues : Array of Extended;

constructor Init(a : PIntArray; b : PExtArray);

procedure GetInputValues(Values : PExtArray);

procedure GetOutputValues;

procedure Excitement;

procedure randomization;

end;

TKohonenLayer = class(TLayer)

public

eta,sigma : Extended;

h,w : Word;

constructor Init(a : PIntArray; b : PExtArray);

procedure Normalize;

function TheWinnerTakesItAll : integer;

function Classic : integer;

procedure Learning(a : integer; SpeedLearn : Extended);

procedure LearningNeib(a : integer; SpeedLearn : Extended);

procedure SigmaInit(s : Extended);

procedure ConvexCombination(delta : Extended);

procedure NeuralGaz(SpeedLearn : Extended);

end;

procedure TKohonenLayer.SigmaInit(s : Extended);

begin

Sigma:=s;

end;

procedure TKohonenLayer.ConvexCombination(delta : Extended);

var i : integer;

begin

eta:=eta+delta;

sigma:=sigma+0.1-10*delta;

for i:=0 to QInputs-1 do InputValues[i]:=InputValues[i]*eta+(1-eta)/sqrt(QInputs);

end;

constructor TKohonenLayer.Init(a : PIntArray; b : PExtArray);

var i : integer;

New : TIntArray;

begin

Inherited Init(a,b);

New:=a^;

H:=New.Value[3];

W:=Round(New.Value[2]/New.Value[3]);

Randomization;

for i:=0 to New.Value[2]-1 do Neurons[i].Pos:=3.75;

eta:=0;

end;

procedure TKohonenLayer.Normalize;

var i : integer;

Sum : Extended;

begin

Sum:=0;

for i:=0 to QInputs-1 do Sum:=Sum+Sqr(InputValues[i]);

for i:=0 to QInputs-1 do

InputValues[i]:=InputValues[i]/Sqrt(Sum);

end;

function TKohonenLayer.TheWinnerTakesItAll : integer;

var i,p : integer;

Min : Extended;

begin

Min:=Neurons[0].Y;

p:=0;

for i:=1 to QNeurons-1 do

begin

if Neurons[i].Pos>0.75 then

if Min>Neurons[i].Y then

begin

p:=i;

Min:=Neurons[i].Y;

end;

end;

for i:=0 to QNeurons-1 do Neurons[i].Y:=0;

for i:=0 to QNeurons-1 do

if i=p then Neurons[i].Pos:=Neurons[i].Pos-0.75

else Neurons[i].Pos:=Neurons[i].Pos+1/QNeurons;

Neurons[p].Y:=1;

GetOutputValues;

TheWinnerTakesItAll:=p;

end;

function TKohonenLayer.Classic : integer;

var i,p : integer;

Min : Extended;

begin

Min:=Neurons[0].Y;

p:=0;

for i:=1 to QNeurons-1 do

begin

if Min>Neurons[i].Y then

begin

p:=i;

Min:=Neurons[i].Y;

end;

end;

for i:=0 to QNeurons-1 do Neurons[i].Y:=0;

Neurons[p].Y:=1;

GetOutputValues;

Classic:=p;

end;

procedure TKohonenLayer.Learning(a : integer; SpeedLearn : Extended);

var i : integer;

begin

for i:=1 to QInputs do

Neurons[a].MassWeight[i]:=Neurons[a].MassWeight[i]+

SpeedLearn*(InputValues[i-1]-Neurons[a].MassWeight[i]);

end;

procedure TKohonenLayer.LearningNeib(a : integer; SpeedLearn : Extended);

var i,j : integer;

begin

for j:=0 to QNeurons-1 do

begin

for i:=1 to QInputs do

Neurons[j].MassWeight[i]:=Neurons[j].MassWeight[i]+

exp(-(Sqr((j div w)-(a div w)) + Sqr((j mod h)-(a mod h)))/(2*Sqr(sigma)))

*SpeedLearn*(InputValues[i-1]-Neurons[j].MassWeight[i]);

end;

end;

procedure TKohonenLayer.NeuralGaz(SpeedLearn : Extended);

var i,j,k,p : integer;

Mass : Array of Extended;

Min : Extended;

begin

SetLength(Mass,QNeurons);

for i:=0 to QNeurons-1 do Mass[i]:=-1;

p:=0;

for i:=0 to QNeurons-1 do

begin

p:=p+1;

Min:=999999;

k:=-1;

for j:=0 to QNeurons-1 do

begin

if Neurons[j].Y<Min then

if Mass[j]=-1 then

begin

k:=j;

Min:=Neurons[j].Y;

end;

end;

Mass[k]:=p;

end;

for j:=0 to QNeurons-1 do

begin

for i:=1 to QInputs do

Neurons[j].MassWeight[i]:=Neurons[j].MassWeight[i]+

exp(-Mass[j]/Sigma)*SpeedLearn*(InputValues[i-1]-Neurons[j].MassWeight[i]);

end;

end;

constructor TNeuron.Init(a : PIntArray; b : PExtArray);

var

New : TIntArray;

begin

New:=a^;

ExtVal:=b^;

IntVal:=TIntArray.Init(2);

IntVal.Value[0]:=New.Value[0]+1;

IntVal.Value[1]:=New.Value[1];

SetLength(MassWeight,IntVal.Value[0]);

if IntVal.Value[0]>0 then MassWeight[0]:=0;

Status:=0;

Y:=0;

Pos:=0;

end;

procedure TNeuron.GetAksonValue(a : PExtArray);

var

i : integer;

b : TExtArray;

begin

b:=a^;

Status:=MassWeight[0];

for i:=1 to IntVal.Value[0]-1 do Status:=Status+MassWeight[i]*b.Value[i-1];

Y:=FunctionActivation;

end;

procedure TNeuron.Distantion(a : PExtArray);

var i : integer;

b : TExtArray;

begin

b:=a^;

Status:=0;

for i:=1 to IntVal.Value[0]-1 do Y:=Y+Sqr(MassWeight[i]-b.Value[i-1]);

Y:=Sqrt(Y);

end;

function TNeuron.FunctionActivation : Extended;

Var m : Extended;

begin

case IntVal.Value[1] of

1 : m:=1/(1+exp(-ExtVal.Value[0]*Status));

2 : m:=ExtVal.Value[0]*Status;

end;

FunctionActivation:=m;

end;

procedure TNeuron.Randomization;

var i : integer;

begin

for i:=0 to IntVal.Value[0]-1 do MassWeight[i]:=random(255);

end;

constructor TLayer.Init(a : PIntArray; b : PExtArray);

var i : integer;

//c : TIntArray;

IntArr : TIntArray;

begin

IntArr:=a^;

QInputs:=IntArr.Value[0];

QNeurons:=IntArr.Value[2];

IntArr.NewLength(2);

ExtArr:=b^;

SetLength(Neurons,SizeOf(TNeuron)*QNeurons);

for i:=0 to QNeurons-1 do Neurons[i]:=TNeuron.Init(@IntArr,b);

SetLength(InputValues,QInputs);

for i:=0 to QInputs-1 do InputValues[i]:=0;

SetLength(OutputValues,QNeurons);

end;

procedure TLayer.GetInputValues(Values : PExtArray);

var i : integer;

a : TExtArray;

begin

a:=Values^;

for i:=0 to QInputs-1 do InputValues[i]:=a.Value[i];

end;

procedure TLayer.Excitement;

var i : integer;

a : TExtArray;

begin

a:=TExtArray.Init(QInputs);

for i:=0 to QInputs-1 do a.Value[i]:=InputValues[i];

for i:=0 to QNeurons-1 do Neurons[i].Distantion(@a);//GetAksonValue(@a);

GetOutputValues;

end;

procedure TLayer.GetOutputValues;

var i : integer;

begin

for i:=0 to QNeurons-1 do OutputValues[i]:=Neurons[i].Y;

end;

procedure TLayer.randomization;

var i : integer;

begin

for i:=0 to QNeurons-1 do Neurons[i].Randomization;

end;

procedure TForm1.Button2Click(Sender: TObject);

var

F : TextFile;

i,j,p,s1,s2,k : integer;

Str : String;

Ch : Char;

Sum : integer;

Temp : Array of String;

begin

OpenDialog1.Filter:='?o?iaeuiue oaee|*.log|';

if (OpenDialog1.Execute) and fileExists(OpenDialog1.FileName) then

begin

AssignFile(F,OpenDialog1.FileName);

Reset(F);

ReadLn(F);

Read(F,Q);

SetLength(Prot,Q);

SetLength(Host,Q);

SetLength(LocalH,Q);

SetLength(Frag,Q);

SetLength(Size,Q);

SetLength(Proc,Q);

SetLength(Active,Q);

SetLength(HACK,Q);

MyList.Clear;

MyList.Add('Iiia?');

MyList.Add('I?ioieie');

MyList.Add('Oino');

MyList.Add('Naiy EAN');

MyList.Add('O?aaiaioaoey');

MyList.Add('?acia?');

MyList.Add('I?ioanni? %%');

MyList.Add('Ioaa?aao');

StringGrid1.RowCount:=Q+1;

StringGrid1.Rows[0]:=MyList;

for i:=0 to Q-1 do

begin

MyList.Clear;

Read(F,j);

MyList.Add(IntToStr(j));

//Memo4.Lines.Add(IntToStr(j));

Read(F,Ch);

Read(F,Ch);

Read(F,Ch);

Prot[i]:='';

While(ch<>' ') do

begin

Prot[i]:=Prot[i]+Ch;

Read(F,Ch);

end;

MyList.Add(Prot[i]);

Read(F,Ch);

Read(F,Ch);

Read(F,Ch);

Host[i]:='';

While(ch<>' ') do

begin

Host[i]:=Host[i]+Ch;

Read(F,Ch);

end;

MyList.Add(Host[i]);

Read(F,Ch);

Read(F,Ch);

Str:='';

While(ch<>' ') do

begin

Str:=Str+Ch;

Read(F,Ch);

end;

MyList.Add(Str);

if Str='YES' then LocalH[i]:=1 else LocalH[i]:=0;

Read(F,Ch);

Read(F,Ch);

Str:='';

While(ch<>' ') do

begin

Str:=Str+Ch;

Read(F,Ch);

end;

MyList.Add(Str);

if Str='YES' then Frag[i]:=1 else Frag[i]:=0;

Read(F,Size[i]);

MyList.Add(IntToStr(Size[i]));

Read(F,Proc[i]);

MyList.Add(IntToStr(Proc[i]));

Read(F,Ch);

Read(F,Ch);

Read(F,Ch);

Str:='';

While(ch<>' ') do

begin

Str:=Str+Ch;

Read(F,Ch);

end;

MyList.Add(Str);

if Str='YES' then Active[i]:=1 else Active[i]:=0;

StringGrid1.Rows[j]:=MyList;

Read(F,Ch);

if Ch='H' then HACK[i]:=1 else HACK[i]:=0;

//Memo4.Lines.Add('**************');

end;

//Memo4.Lines.Add(IntToStr(Q));

CloseFile(F);

SetLength(Temp,10);

SetLength(Mass,(Q-1)*11);

SetLength(SHack,Q-1);

for i:=0 to Q-12 do

begin

Mass[12*i]:=0;

for j:=0 to 9 do Mass[12*i]:=Mass[12*i]+LocalH[i+j];

Mass[12*i+1]:=0;

for j:=0 to 9 do Mass[12*i+1]:=Mass[12*i+1]+Frag[i+j];

Mass[12*i+2]:=0;

Mass[12*i+3]:=0;

for j:=0 to 9 do

if Prot[i+j]='TCP' then Mass[12*i+2]:=Mass[12*i+2]+1;

for j:=0 to 9 do

if Prot[i+j]='UDP' then Mass[12*i+3]:=Mass[12*i+3]+1;

Sum:=1;

s1:=0;

for j:=0 to 9 do

begin

Str:=Host[i+j];

p:=0;

for k:=0 to 9 do

begin

if Str=Host[i+k] then

begin

p:=p+1;

s2:=k;

end;

end;

if p>Sum then

begin

Sum:=p;

s1:=s2;

end;

end;

Mass[12*i+4]:=Sum;

Mass[12*i+5]:=LocalH[i+s1];

Sum:=0;

for j:=0 to 9 do Sum:=Sum+Proc[i+j];

Mass[12*i+6]:=Sum/10;

Mass[12*i+7]:=Proc[i+9]-Proc[i];

Sum:=0;

for j:=0 to 9 do Sum:=Sum+Size[i+j];

Mass[12*i+8]:=Sum/10;

Sum:=0;

for j:=0 to 9 do

if (Size[i+j]>=0.8*Mass[12*i+8])and

(Size[i+j]<=1.2*Mass[12*i+8]) then Sum:=Sum+1;

Mass[12*i+9]:=Sum;

Sum:=0;

for j:=0 to 9 do Sum:=Sum+Active[i+j];

Mass[12*i+10]:=Sum;

for j:=0 to 9 do Temp[j]:=Host[i+j];

for j:=0 to 8 do

begin

Str:=Temp[j];

for k:=0 to 9 do

if k<>j then

if Str=Temp[k] then Temp[k]:='';

end;

Sum:=0;

for j:=0 to 9 do

if Temp[j]<>'' then Sum:=Sum+1;

Mass[12*i+11]:=Sum;

Sum:=0;

for j:=0 to 9 do Sum:=Sum+HACK[i+j];

SHAck[i]:=Sum;

end;

end;

end;

procedure TForm1.InitializationMap;

var i,j,p : integer;

begin

for i:=0 to H*W-1 do

begin

p:=Random(Q-13);

for j:=0 to 10 do

KMap.Neurons[i].MassWeight[j+1]:=Mass[12*p+j];

end;

end;

procedure TForm1.Button3Click(Sender: TObject);

var i,j,p,k,m : integer;

Quant,Winner : integer;

Part : TExtArray;

SMax,SMin : Extended;

begin

InitializationMap;

for i:=0 to Image1.Height-1 do

for j:=0 to Image1.Width-1 do

Image1.Picture.Bitmap.Canvas.Pixels[i,j]:=RGB(150,150,150);

Quant:=StrToInt(Edit3.Text);

Part:=TExtArray.Init(12);

SMax:=0.7;

SMin:=0.0001;

ProgressBar1.Max:=Quant;

ProgressBar1.Position:=0;

for i:=0 to Quant-1 do

begin

KMap.SigmaInit(10*(1-i/Quant)+0.1);

p:=Random(Q-12);

for j:=0 to 11 do Part.Value[j]:=Mass[12*p+j];

KMap.GetInputValues(@Part);

KMap.Excitement;

Case RadioGroup1.ItemIndex of

0:

begin

Winner:=KMap.Classic;

if CheckBox1.Checked then KMap.LearningNeib(Winner,SMax-(SMax-SMin)*i/Quant)

else KMap.Learning(Winner,SMax-(SMax-SMin)*i/Quant);

end;

1:

begin

Winner:=KMAp.TheWinnerTakesItAll;

if CheckBox1.Checked then KMap.LearningNeib(Winner,SMax-(SMax-SMin)*i/Quant)

else KMap.Learning(Winner,SMax-(SMax-SMin)*i/Quant)

end;

2:

begin

KMap.NeuralGaz(SMax-(SMax-SMin)*i/Quant);

end;

end;

ProgressBar1.StepBy(1);

end;

ProgressBar1.Position:=0;

for i:=0 to KMap.QNeurons-1 do

KMap.Neurons[i].MassWeight[0]:=0;

for i:=0 to Q-12 do

begin

for j:=0 to 11 do Part.Value[j]:=Mass[12*i+j];

KMap.GetInputValues(@Part);

KMap.Excitement;

Winner:=KMap.Classic;

KMap.Neurons[Winner].MassWeight[0]:=1;

//Memo4.Lines.Add(IntToStr(Winner));

if SHack[i]>=8 then

begin

for m:=0 to W1-1 do

for k:=0 to H1-1 do

Image1.Picture.Bitmap.Canvas.Pixels[(Winner div W)*W1+m,(Winner mod W)*H1+k]:=RGB(255,0,0);

end

else if SHack[i]=7 then

begin

for m:=0 to W1-1 do

for k:=0 to H1-1 do

Image1.Picture.Bitmap.Canvas.Pixels[(Winner div W)*W1+m,(Winner mod W)*H1+k]:=RGB(255,40,40);

end

else if SHack[i]=6 then

begin

for m:=0 to W1-1 do

for k:=0 to H1-1 do

Image1.Picture.Bitmap.Canvas.Pixels[(Winner div W)*W1+m,(Winner mod W)*H1+k]:=RGB(225,80,80);

end

else if SHack[i]=5 then

begin

for m:=0 to W1-1 do

for k:=0 to H1-1 do

Image1.Picture.Bitmap.Canvas.Pixels[(Winner div W)*W1+m,(Winner mod W)*H1+k]:=RGB(225,120,120);

end

else if SHack[i]=4 then

begin

for m:=0 to W1-1 do

for k:=0 to H1-1 do

Image1.Picture.Bitmap.Canvas.Pixels[(Winner div W)*W1+m,(Winner mod W)*H1+k]:=RGB(225,160,160);

end

else if SHack[i]=3 then

begin

for m:=0 to W1-1 do

for k:=0 to H1-1 do

Image1.Picture.Bitmap.Canvas.Pixels[(Winner div W)*W1+m,(Winner mod W)*H1+k]:=RGB(225,200,200);

end

else

begin

for m:=0 to W1-1 do

for k:=0 to H1-1 do

Image1.Picture.Bitmap.Canvas.Pixels[(Winner div W)*W1+m,(Winner mod W)*H1+k]:=RGB(225,225,225);

end;

//Image2.Picture.Bitmap.Canvas.

//.Pixels[j,i]:=RGB(

end;

for i:=0 to KMap.QNeurons-1 do

begin

if KMap.Neurons[i].MassWeight[0]=1 then

begin

Memo3.Lines.Add('Iae?ii '+IntToStr(i));

for j:=0 to KMap.QInputs-1 do

Memo3.Lines.Add(FloatToStr(KMap.Neurons[i].MassWeight[j]));

end;

end;

end;