Edify Friday: Parallel Computing in .NET 4.0 with PLINQ

The Institute for Cyber Security uses C# and the .NET Framework for our startups in incubation, internal research projects, community exercises, and more. With the second beta of Visual Studio 2010 out, we've been really interested in the framework and language changes, and are evaluating it for integration into our software process. One change in particular is PLINQ, which stands for Parallel Language Integrated Query. This builds on LINQ, introduced two years ago, which added native data querying capabilities to .NET languages. Thus, today's Edify Friday covers a simple way to use PLINQ in C# code.

First, let's take a look at a simple LINQ example. Below is a simple program that counts the number of even numbers up to a ceiling (see Listing 1 below):

using System;
using System.Linq;
using System.Collections.Generic;
using System.Diagnostics;

namespace PlinqExample {
  class HelloWorldPlinq {
    static void Main(string[] args) {
      Stopwatch sw = Stopwatch.StartNew();
      int numEven = CountEven(1000000);
      Console.Write("Found {0} even numbers. ", numEven);
      Console.WriteLine("Elapsed: {0} ms", sw.ElapsedMilliseconds);
    }

    private static int CountEven(int top) {
      IEnumerable<int> allNumbers = Enumerable.Range(1, top);
      var q = from i in allNumbers where (i % 2 == 0) select i;
      return q.Count();
    }
  }
}

Listing 1: HelloWorldPlinq.cs

When running the application, one should get something similar to the following: "Found 500000 even numbers. Elapsed: 18 ms." (naturally, the elapsed time will vary based on CPU power and processor availability).

Now, we'll modify one line from above to make use of PLINQ (see Listing 2 below):

private static int CountEven(int top) {
  IEnumerable<int> allNumbers = Enumerable.Range(1, top);
  var q = from i in allNumbers.AsParallel() where (i % 2 == 0) select i;
  return q.Count();
}

Listing 2: Partial Listing of HelloWorldPlinq.cs

By using the AsParralel() method, we effectively cast q as a different object entirely that can take advantage of threading for its calculations. Now running the application may give you something similar to the following: "Found 500000 even numbers. Elapsed: 14 ms."

Got a topic that you'd like to see covered in Edify Friday? Leave a comment below or email me!

Erhan J Kartaltepe, PMP
erhan.kartaltepe-at-utsa.edu

ICS in the News: ICS Advises the Air Force Association on its Cyber Defense Competition

Reuters had an interesting article on the operations division at ICS, covering the Air Force Association's Cyber Defense Competition. In the article, the AFA mentions that competition benefits greatly through the advice and assistance of ICS's Greg White, director of ICS's operations division. Through Greg's efforts, ICS also created and conducts the National Collegiate Cyber Defense Competition.

Additionally, The Journal covered the announcement as well, covering how the 125,000-member AFA is geared to promote public understanding of aerospace power and its critical relationship to national security. The article also describes Greg White as a "cyber security expert" and discusses his service as an advisor for the competition.

For more information on ICS and its operations, commercialization, and research divisions, visit the ICS site.

Erhan J. Kartaltepe, PMP
erhan.kartaltepe-at-utsa.edu

Cyber Security Awareness: Free Training and Webcasts

David Pickens posted an article about free cyber security awareness training and webcasts. This is quite appropriate considering October is Cyber Security Awareness Month.

While all are great and each of the eight organizations listed we at the Institute for Cyber Security couldn't help but notice one in particular, highlighting our operational division. From the article: "The operational division of the UTSA Institute for Cyber Security...is designed to leverage its infrastructure assurance and security strengths as part of the solution to the nation's homeland security needs."

To learn more about the training and webcasts, check out the article. For more information on ICS and its operations, commercialization, and research divisions, visit the  ICS site.

Erhan J. Kartaltepe, PMP
erhan.kartaltepe-at-utsa.edu

Edify Friday: Death Star Threat Modeling

The Institute for Cyber Security, as one may suspect, deals with security issues on a daily basis. Whether it's with our startups in incubation, internal research projects, presentations at conferences, or something else entirely, security vocabulary is used to describe very specific situations. When using terms like "vulnerability" or "threat", which have precise meanings in an information security context, it helps when everyone is on the same page with the vocabulary. Thus, today's Edify Friday covers threat modeling at a high level, with an object lesson at the end:

• An asset is a resource of some value. This value could be real or perceived, and could be different from the owner and the attacker.
• A vulnerability is a weakness that could be exploited to harm the asset. This could be unintended, like a software bug, or intended, such as an assumption that the vulnerability is unlikely to be exploited.
• A threat is something that can harm the asset, usually by exploiting the vulnerability. The intent of the threat is irrelevant, and could be man-made, automated, or natural.
• A risk is the likelihood that the harm will occur. Risk is present in any system, and is a function of an asset's value, the vulnerability's severity, and threat's probability.
• A countermeasure is a control that reduces risk to some level. This must be balanced against the risk probability and the asset's value.

Put another way, risk is the chance that a threat will attack or exploit a vulnerability. My favorite view of this is the "Death Star Threat Model" (shamelessly translated from Kevin William's presentation last year at the Last Hope conference). Here, the asset was the Death Star; the vulnerability of the Death Star was the two meter-wide thermal exhaust port that's shaft lead directly to the reactor system; the threat to the Death Star was the Rebel Alliance's X-wing star-fighters; the risk was the slim chance that a small one-man fighter could penetrate the outer defenses, use a proton torpedo to score a precise hit, and start a chain reaction that will destroy the battle station; and the countermeasure was a deployment of Imperial tie fighters around the Death Star.

That's threat modeling in a nutshell.

Got a topic that you'd like to see covered in Edify Friday? Leave a comment below or email me!

Erhan J Kartaltepe,
erhan.kartaltepe-at-utsa.edu

The Difference Between C#'s "as" and "cast" Operators

 

Eric Lippert's blog had an interesting article today on the difference between C#'s 'as" and "cast" operators. As he states in the opening paragraph: "Most people will tell you that the difference between '(Alpha) bravo' and 'bravo as Alpha' is that the former throws an exception if the conversion fails, whereas the latter returns null. Though this is correct, and this is the most obvious difference, it's not the only difference."

He then walks through some of the minutia between the two operations, which makes for a short interesting read for developers interested in the semantics of the language.

Image is of Harry Potter using the cast operator.  

Erhan J. Kartaltepe,
erhan.kartaltepe-at-utsa.edu

Edify Friday: Python and .NET Interoperability

At the Institute for Cyber Security, we use develop mobile applications and use Python and .NET Framework for our startups in incubation or in‐house projects, and every once in a while, we need to have them interoperate.

Today's Edify Friday covers how simple it is to use .NET code, either in the framework or in-house, in a Python application. It makes use of the Python ctypes library, allowing one to call functions in DLLs. In the example we walkthrough, we call the MessageBoxA function from WinDLL.

First, we "import" the libraries and Windows types we need. The function prototype WINFUNCTYPE is Windows-only; there are others for Python and C. Next,for ease of use, we're going to create a prototype of the MessageBoxA function, where the first parameter is the return type (MessageBoxA returns an integer) and the remaining parameters mirror MessageBoxA's parameters.

After we create the prototype, we can set named parameters in the usual Python style, giving default values for them if we like. Finally, we'll save this prototype to a variable (think delegates from .NET) we call MessageBox, giving it the class MessaBoxA belongs to, and its parameters.

The example below puts it all together in Python 3.1:

Figure 1: Windows Message Box in Python

Got a topic that you'd like to see covered in Edify Friday? Leave a comment below or email me!

Erhan J. Kartaltepe,
erhan.kartaltepe-at-utsa.edu

Cyber Security Awareness Week at the Polytechnic Institute of New York University

It's Cyber Security Awareness Week (CSAW) at the Polytechnic Institute of New York University. This seems like as good a week as any to raise awareness of cyber security, so the Institute for Cyber Security shall in spirit celebrate this week as well. :-)

Last year, 500 students competed in these challenges, and their sixth annual CSAW looks to trump that--they've alread taking registration for their annual competition, which will unfold over the next couple of months. The challenges pit high school, college, and graduate students against each other in a series of cyber-security-related competitions, which includes analyzing an infected hardware platform, detecting application security vulnerabilities, and solving a fictional murder mystery using digital forensics.

Their main site and Facebook page has more information on their contest, and prizes include cash and scholarships, as well as travel grants. This is a great idea and looks like a fantastic way to engage students in studying or even pursuing cyber security. Best of luck to all participants!

Erhan J. Kartaltepe,
erhan.kartaltepe-at-utsa.edu

The Four Types of Aspect-Oriented Programming (AOP) Weavers

At the Institute for Cyber Security, we use many different technologies to help accomplish our goals. One such technology is Aspect-Oriented Programming (AOP)., where one can "weave" cross-cutting concerns into source code, separating them from core concerns.

While compile-time weaving is common, the weaving process itself can take place at one of four different times: compile-time, post-compile time, load-time, and run-time. The run-time behavior of an application are the same regardless of the approach chosen.

• Compile-time weaving: this is the simplest approach. A compile-time weaver will compile from source and produce woven class files as output. The aspects themselves may be in source or binary form.
• Binary-weaving (also called post-compile-time weaving): this is used to weave existing class files and JAR files. As with compile-time weaving, the aspects used for weaving may be in source or binary form, and may themselves be woven by aspects.
• Load-time weaving: here, binary weaving is defered until the point that a class loader loads a class file and defines the class to the system. To support this, one or more "weaving class loaders", either provided explicitly by the run-time environment or enabled through a "weaving agent" are required. 
• Run-time weaving: this is the weaving of classes that have already been defined to the system without reloading those classes, making it a more dynamic load-time weaver.

Erhan J. Kartaltepe,
erhan.kartaltepe-at-utsa.edu

The Basics of Base64

At the Institute for Cyber Security, much of our work with our startups in incubation or research involves transmitting data using Base64. Whether in the form of files or images, data between machines with incompatible data definitions, or encrypted information, Base64 is used almost as a de facto standard for content transfer encoding.

Almost any high-level language will has a library that will encode and decode data from Base64 and back, but it's worth knowing the underlying principle behind the transform. The way it works is simple: whether text, image, or encrypted information, every bit of data is a number stored in memory. By reading that data in blocks of six bits (2⁶ = 64) and mapping each number to a character (0 = 'A', 1 = 'B', and so on, until 62 = '+' and 63 = '/'; see the figure up above), binary data of any size can be encoded into a string. See Figure 1 for an example, taking a string to its constituent bytes, then reading each group of six bits into a new Base64 character:

Figure 1: Encoding "Hello!" in Base64

Erhan J. Kartaltepe,
erhan.kartaltepe-at-utsa.edu

Implementing the Decorator Pattern in AspectJ

Last week we covered how we'd implement the decorator pattern in Java to solve how to modify the behavior of an existing class without modifying or subclassing it. It turns out that implementing this behavior is even simpler in AspectJ.For a refresher on the decorator pattern, read the preceding post.

As before, say we have a Car class which has some functionality we'd like to modify, say by adding a top to the car or an engine. Listing 1 shows a Car interface and SimpleCar class.

Listing 1: Car Interface and SimpleCar Class

To add a top or an engine, we'd need separate classes for each functionality, and an abstract class that they extend. But instead, we can just add two aspects (see Listing 2).

Listing 2: TopAspect and EngineAspect Aspects

Each aspect modifies the code so that it performs some behavior prior to calling drive(), and it modifies the getInfo method. Both aspects total twelve lines of code (not including braces), whereas the equivalent code in Java totals nearly double the lines at twenty-two.

We can view the results of the Decorator pattern by calling the getInfo method, which should return "car, with top, with revved engine".

Listing 3: DecoratedTest Class

Line three is very straightforward, whereas the equivalent line in the Java-only implementation is a bit more convoluted, as each constructor must be called for each class. For multiple classes, it can be quite unwieldy.

This is a bit perfect example of using AspectJ's interception capabilities to add a crosscutting concern. In a more realistic example, Car could be a communication class, and TopAspect could be a logger, while EngineAspect could be an encryption engine. In this way, Car could be written without "concern" for logic unrelated to itself, and these new aspects could be added later.

Erhan J. Kartaltepe,
erhan.kartaltepe-at-utsa.edu