Monday, April 10, 2017

Performance of MD5, SHA1, SHA256, SHA384, and SHA512 with C#

Recently I ran a few performance calculations to see the differences of performance of MD5, SHA1, SHA256, SHA384, and SHA512. I ran the tests on my box with one code set. I compiled my code as a 32-bit program and as a 64-bit program. Besides the performance of each hashing algorithm, I wanted see the difference in 32-bit and 64-bit compiles would make.


My testing method consisted of the following method.
  • CPU: Intel(R) Core (TM) i7-4700MQ CPU 2.40GHz
  • OS: Windows 7, 64-bit Operating System
  • Memory: 32.0 GB (31.6 GB usable)
  • Hash string size was 1K.
  • Ran in release mode, (not inside of Visual Studio, language used was C#)
  • Each hash string method was called 10,000 times with a different string to be hashed.
  • Time was accumulated in CPU ticks.
  • I ran each the program three times.
 

The one thing that stood out to me was SHA-256. I was surprised by the performance, the small difference between 32 and 64-bit programs and the actual time it took to create an SHA-256 hash. After spending some time on the internet I came up with the following.
  




 
SHA-256 algorithm generates an almost-unique, fixed-size 256-bit (32-byte) hash. So even on a 64-bit machine where the word size is larger to push more data thru the transformations algorithm because SHA-256 is using a fix 32-byte data you do not see the larger difference between a 32 and 64 bit processors.  One interesting note is Microsoft has not updated its core base class from which all implementations of cryptographic hash algorithms. The HashAlgorithm class is still is using 32-bit transformblocks. I am curious why Microsoft hasn't change to use a 64-bit transformblock. Would that improve the performance level? I would have thought Microssoft would have a 64-bit version of SHA256.




The largest takeaway here is "Do not use MD5 or SHA1". The difference in performance isn't going to hurt your application as much as using an outdated hashing algorithm will.



Code.

  Calling method to create a Hash string. All hashing followed the same template.


         static Random random = new Random();
         static Int64 MaxIterations = 10000;
         static Int16 TotalHashLength = 1024;
              static void MD5()
              {
                  MD5Hash md5Hash = new MD5Hash();
                  for (int i = 0; i < MaxIterations - 1; i++)
                  {
                      md5Hash.CalculateHash(RandomString(TotalHashLength));
                  }
                  file.WriteLine("MD5 , {0}", md5Hash.TotalElapseTime().ToString());
              }
 

 Each time an hash was created a different random string was created using lower, upper case, number and special printable characters.

 
              static string RandomString(int length)
              {
                  const string chars = "Aa0BbCcDd1+EeFf2_GgHh3(IiJj4)KkLl5*MmNn6&OoPp7^QqRr8%SsTt9#UuVv@WwXxYyZz!";
                  StringBuilder sb = new StringBuilder();
                  for (int i = 0; i < length; i++)
                  {
                      sb.Append(new string(Enumerable.Repeat(chars, 1).Select(s => s[random.Next(s.Length)]).ToArray()));
                  }
                  return sb.ToString();
             }
   

 Main hashing calculation for MD5, other hashing methods follow this template.


              private Stopwatch sw;
              private MD5 md5;
              private TimeSpan ElapseTime;
    
               public void CalculateHash(string input)
               {
                  sw = Stopwatch.StartNew();
                  md5 = MD5.Create();
                  byte[] inputBytes = System.Text.Encoding.ASCII.GetBytes(input);
                  byte[] hash = md5.ComputeHash(inputBytes););
                  ElapseTime += sw.Elapsed;
                  sw.Stop();
               }




For the complete code solution see my github page https://github.com/lwconklin/Hashing