private void NetAnalyze(object obj)
{
OperationContext context = (OperationContext)obj;
while (true)
{
if (netQueue.Count > 0)
{
string[] packets;
lock (netQueue)
{
packets = netQueue.Dequeue();
}
foreach (string packet in packets)
{
string[] packetInfoArray = packet.Split('|');
double[] input = Utilities.ParseToNetUnit(packetInfoArray[4], DELIMETER);
if (!netClassifier.Classify(input))
{
context.GetCallbackChannel <IAnalyzerCallback>().GenerateNetWarning(packetInfoArray);
}
}
}
else
{
Thread.Sleep(5000);
}
}
}
static void Main(string[] args)
{
int ecount = 0, fcount = 0;
double[,] sampling = Utilities.ReadNetTrainingFile(@"E:\Диплом\WorkingDirectory\training.txt", ref ecount, ref fcount);
SimpleClassifierNN classifier = new SimpleClassifierNN(sampling, 36, 6000, 10, 1500);
classifier.Train();
classifier.SaveNetwork(String.Format(@"E:\Диплом\WorkingDirectory\netSave.txt", Directory.GetCurrentDirectory()));
int i = 0;
using (System.IO.StreamReader file = new System.IO.StreamReader(@"E:\Диплом\WorkingDirectory\test1.txt"))
{
string line;
while ((line = file.ReadLine()) != null && line != "")
{
string[] allNetEntry = line.Split('|');
double[] input = Utilities.ParseToNetUnit(allNetEntry[4], ';');
string desc = "";
Console.WriteLine(String.Format("Result{0}: {1}", i, classifier.Classify(input).ToString()));
i++;
}
}
string infile = @"E:\Диплом\WorkingDirectory\SecurityTraining.txt";
string outfile = @"E:\Диплом\WorkingDirectory\logOutput.txt";
string testfile = @"E:\Диплом\WorkingDirectory\logTest.txt";
string saveFile = @"E:\Диплом\WorkingDirectory\logSave.txt";
LogClassifier logcl = new LogClassifier(saveFile, infile);
var test = Utilities.ReadHostClassifyFile(testfile);
logcl.SaveClassifier(saveFile);
CheckHostPackets(test[0], logcl);
}
public double[] CreateNewNN(string trainingFileName, string testFileName, string goal, int epochCount, int neuronCountInHiddenLayer)
{
List <double> results = new List <double>();
int trainingSamples = 0, countFeatures = 0, testSamples = 0, testGood = 0, testBad = 0, TP = 0, TN = 0, FP = 0, FN = 0;
temporaryGoal = goal;
if (goal == "NET")
{
temporaryNetClassifier = null;
var trainingData = Utilities.ReadNetTrainingFile(trainingFileName, ref trainingSamples, ref countFeatures);
if (trainingData == null)
{
return(null);
}
temporaryNetClassifier = new SimpleClassifierNN(trainingData, countFeatures, trainingSamples, neuronCountInHiddenLayer, epochCount);
var trainingResult = temporaryNetClassifier.Train();
var testData = Utilities.ReadNetTestFile(testFileName, ref testSamples);
foreach (var unitTest in testData)
{
bool classifyResult = temporaryNetClassifier.Classify(unitTest.Item1);
if (unitTest.Item2)
{
testGood++;
if (unitTest.Item2 != classifyResult)
{
FN++;
}
else
{
TP++;
}
}
else
{
testBad++;
if (unitTest.Item2 != classifyResult)
{
FP++;
}
else
{
TN++;
}
}
}
}
if (goal == "HOST")
{
temporaryHostClassifier = null;
temporaryHostClassifier = new LogClassifier(trainingFileName, neuronCountInHiddenLayer, epochCount);
var testData = Utilities.ReadHostFile(testFileName, ref testSamples);
for (int i = 0; i < testData.Item1.Length; i++)
{
bool classifyResult = temporaryHostClassifier.TestAnalyze(testData.Item1[i]);
if (testData.Item2[i])
{
testGood++;
if (testData.Item2[i] != classifyResult)
{
FN++;
}
else
{
TP++;
}
}
else
{
testBad++;
if (testData.Item2[i] != classifyResult)
{
FP++;
}
else
{
TN++;
}
}
}
}
double precision = (double)TP / (double)(TP + FP);
double recall = (double)TP / (double)(TP + FN);
double accuracy = 2 * ((double)(precision * recall) / (double)(precision + recall));
double firstMistake = (double)FP / (double)testSamples;
double secondMistake = (double)FN / (double)testSamples;
results.Add(trainingSamples);
results.Add(testGood);
results.Add(testBad);
results.Add(testSamples);
results.Add(epochCount);
results.Add(accuracy);
results.Add(firstMistake);
results.Add(secondMistake);
return(results.ToArray());
}