A while back some folks over at Microsoft put together a draft of something they call the "Secure Development Lifecycle". This lifecycle consists of several suggested steps an organization can go through to help encourage secure development.
You might come across the term "push left" when talking about secure development. Traditionally, security validation and exploit location has been a step further along in the process than development - usually something QA does. The "push left" mentality puts development to the left of QA/testing in the flow of the lifecycle and advocates the integration of these practices early.
There's five main steps to the actual workflow that makes up their suggested process:
There's two other steps that are a bit "outside" of the rest of this flow but should still be considered very much a part of the process - the training of the development group and the creation of an incident response plan when issues come up.
Now, there's been whole books written on this flow, so I'm not going to get into too much detail on it. I just want to give you a high level look at each of the sections so it can get the wheels turning in your own mind about making it work in your environment.
At the start of any software project, there's always a requirements gathering phase. Even if you're following some of the agile methodologies, you still have to have something to write stories about. Coming up with the security standards and requirements for your application is no different. These should be gathered by the same people, in fact. Since the idea of secure development is to start at the very beginning of the coding, it helps to have both the feature requirements and the security requirements side by side during planning.
Since application security is such a subjective thing to think about, there's no hard and fast rules about what might need to be included. Here's a few things to think about, though, to help you along your way:
Obviously, this is just a small sampling of things to think about. There'll be a lot more that will be dependent on your application type. Be sure you think about all aspects of the data flowing through your system and how your users will interact with it.
Be sure to think about the technologies you're planning to involve and how much of them involve third party (possibly untested) code. Look for libraries that have been around for a while and seem well developed. If you've been developing PHP for any length of time you can usually look over the code of a library and tell if the author knows what they're doing. Look on Github for projects that have a good history of resolving issues, are full featured and have some sort of community around them. There's a growing trend in PHP development of unit testing. Look for something that has these tests as it can tell you a lot about the stability of the project and how much dedication they have to the end result.
I> This is also a good time to define the security standards documentation that developers I> on the project should follow as they progress through the application. Give them a I> "checklist" of sorts they can keep handy and refer to often.
Another thing to think about here is the tools you're going to use to enforce the standards you're setting up. Unfortunately, there's not too many PHP-centric tools out there to help with this kind of thing. There's one tried-and-true method that's proven to not only create better code but also more secure code - code reviews. I hear of more and more organizations that are implementing code reviews as a part of their standard development practices. Basically, a code review (sometimes called a "peer review") is another developer looking over the code you're committing and ensuring it's up to the defined standards.
PHP have something that makes the automated checking of the code style simpler with PHP_CodeSnffer, but there's not much out there for security testing. There are some static scanning tools that can help to catch unsafe practices like using exec or directly outputting data without filtering it, but they're not very well maintained. You can also get this as a service from another company like Veracode.
It's also recommended that you come up with the "bug bars" or levels of allowed bugs and types that can be in the different stages fo your development. Several groups will divide this up by environment. For example, they'll say that, in order for a release to move on to be tested by QA, there has to be five or less bugs in the development environment. This also sets the level of these bugs - say, of the five, only one can have a "High" severity and the rest must be "Medium" or below.
This part of the process isn't necessarily something that developers need to do, but it can't hurt to be involved. In the following chapter I talk some about the idea of "risk" and how to determine it for you application. There's a lot involved in trying to estimate risk, and it helps to think about those things as you write your code. If you can get in on some of the threat modeling and risk evaluations done for the application - even if it's just once - it'd be beneficial.
Once you have the requirements firmly in hand you can start on the fun part - actually planning out the design of the system. Ask any developer and they'll tell you that getting into the technology choices and structure of the application is one of their favorite parts. If you've been lucky enough to have been given a "green field" project (brand new) you're doubly lucky. There's no legacy code to have to worry about, just a wide open space where you can start putting stakes in the ground.
This is the part of the process where you start thinking technically. You look at the requirements you've been given and try to figure out a good structure for the overall application and which parts are going to be the most important security-wise. If there's an architect as a part of your team, it'll fall in his or her lap to think about these things and give a "game plan" to the rest of the development staff for future development.
An "attack surface" is something that might be new to some developers out there. The basic idea is that you take a look at your application from the outside and think about what's exposed that a would-be attacker could abuse. This could include things like:
It's up to you, as a developer, to think about minimizing this "surface". Following principles like the "least privilege" and "defense in depth" can help this. For more information on these, check out the "Best Practices" chapter near the end of this book.
Threat modeling is the process by which you would go through the parts of your application and think about the threats that could effect the functionality. If you've never done this kind of thing before it can be a little daunting. There are, however, some frameworks out there to help guide you through the right kind of thinking.
STRIDE is another Microsoft invention. It's an acronym that stands for Spoofing identity, Tampering with data, Repudiation, Informaton disclosure, Denial of Service and Elevation of privilege. Using the STRIDE model, you can look through your features and evaluate the level of risk that's associated with each of these options. By breaking up your application into bite sized chunks, you can more effectively run it past these common threats.
Don't be fooled into thinking that worrying about the risks of each piece individually it enough. Most software does its work through components interacting with each other, so be sure that you take into account how the components will talk to the others when assessing risk.
Another common approach is DREAD modeling. DREAD is another acronym that stands for Damage, Reproducabiity, Exploitability, Affected users and Discoverability. This approach is a bit different than the STRIDE way of thinking. With STRIDE, you have a specific set of threats to think about and help resolve. With DREAD it's more about the end result of the attack than the method used. There's a focus on how the attack effects the system overall and the difficulty of finding and reproducing the attack is.
With the planning of that "what" and "how" you're building the system with out of the way, you can get down to the business of actually implementing features. With a secure coding standard firmly in hand, you can start to talk that first set of stories (you are doing agile, aren't you?) and start building up the system and its structure. All of the tools and techniques that you planned out in the previous steps are put to use here. Approved tools, methods for testing and coding standards should be applied and tested for as the development progresses.
During this time, some effort should be spent to determine what kinds of situations and
functionality of PHP you want to avoid. Generally, there's some pretty easy wins here
by restricting things like eval, exec and
its cousins as well as things like the direct use of the MySQL extensions (PDO and bound
parameters is a much better option). You can use the
disable_functions configuration setting
php.ini to turn off the usage of those methods as a means of enforcing that
part of the coding standard. Obviously, there's going to be certain situations where
some of those methods are needed, but those should be handled on a case by case basis.
You should also think about the versions of the tools and libraries that you're using. Try to keep things up to date and as patched as possible.
This does not necessarily mean using the latest and greatest versions. Often those early adopters find themselves having to revert back to a previous version because of security concerns or functional issues. PHP itself isn't even immune to this kind of thing. There's been several occasions where the PHP development group has had to push out a new version a day or two after a previous release just to fix security issues.
Track the status of the libraries you're using and, if need be, lock them down to a specific version until you feel comfortable enough to update. Things like Composer make this simpler by offering the ability to define the tag or branch you want to pull the library from on Github.
As development it progressing, according to the agile methodologies, the code that's coming out from the developers should be continuously tested. Sometimes this means humans sitting in front of a keyboard and mouse clicking away on the application, but a lot of times it comes in the form of automated testing. As was mentioned before, there's not a lot of PHP-centric testing tools out there, but there are plenty of web application-related tools that can use and abuse your app regardless of the underlying technology.
This kind of testing is knowns as "dynamic analysis" and can be run with any number of scanners, both commerical and free for use. Here's some of the more popular free/open source tools out there:
Point these tools at the frontend of your application and set them to work. They'll produce reports you can use to find some of the flaws in your applications before the attackers do. I still haven't found a good one that will output a parsable report (maybe something in XML or even just a text file) but some of them will return an response code from their execution indicating the pass/fail status.
If you don't want to go the scanner route and feel like some more in-depth testing is needed, you can always go with something like Seleinum or Behat for frontend testing. Behat is a PHP-based behavior-driven development tool that uses tests defined in a Gherkin syntax for its tests, making them as readable as plain English. Here's an example:
Feature: Testing the user auth for a correct response Scenario: Test with a valid username/password combo Given that I am on "/user/login" When I fill in "username" with "testuser1" When I fill in "password" with "testpass1" And I press "loginButton" Then I should see "Login Success!"
This, along with a good set of unit tests (with something like PHPUnit), can be executed as often as needed. Most development groups will make test execution a part of their continuous integration efforts. They run these and other tasks via something like Jenkins to track the status of the quality of their software.
Having a good automated testing setup and some good human-based exploratory testing is good, but when there's major changes in your application like the addition of a new library or a new section of the site. This should trigger a reassessment of the threat assessment of the application and a reevaluation of it's attack surface. Software may be made up of individual modules, but those modules have to work together to make things happen. As as result, a change in one part of the application could effect a number of systems and this impact needs to be tested.
This usually involves less testing on the automated side and more of the human testers executing their own scripts and running through their own plans to ensure the functionality of the application overall. Specific focus should be given to features that directly implemented the changed code.
Thankfully as developers, being able to execute our own automated tests can help reduce the number of issues that QA might come across. This holds especially true for an existing codebase with a good set of tests (unit and functional). Having a well-tested application allows you to "recklessly refactor" and still be sure your application will function as expected.
Finally the day has come to unleash the hard work you've been doing and set it free for customers to use and abuse. There's just a few things to clean up before you push that release out the door, though. You want to be sure you have a few things checked off the list:
A plan for handling the issues that come in (yes, there'll be issues - no code is perfect) This kind of plan can involve not only setting up certain people to handle the reports of the issues but also the tools you'll need to manage the issues that come up. Usually this will be a bug tracker with a specific type of issue for user-reported issues. By keeping them along side the rest of the issues the development group deals with, they can easily be prioritized with the rest of the work.
The QA group should run the application through one final security review before blessing it and sending it down the line. This review should be done on an environment that's as close to the end user system or, if possible, multiple environments at once to ensure that there's no cross-platform issues. If all goes well, the application will pass with flying colors and you can call it good.
Of course, this is the real world and we all know there'll be something that comes up despite through testing through out the entire process. These issues can either be taken as "acceptable risks" to be fixed later or they might be "show stoppers" and result in a dash to fix issues prior to the release. Hopefully you'll have enough time before the final release date of the app to have a little slack to fix these kinds of issues.
Finally, when the final review is done and everyone's signed off on everything you're ready to push it out the door and see how it does. Be sure that you archive everything you've done so far and document everything that needs to be done, what was put off as a "Should" or "Pending" issue as well as any documentation that might go with the release. This also includes all of the tests, both the automated and manual processes, that was used to ensure the app was working as it should. The idea is that you can have something to look back at and know the state of the software at that time.
Developers have a bit easier time with this as most of their stuff should be in source control anyway.
Entire books have been written about this suggested process. If you're interested in more detail, I'd suggest checking out some of the links below for more information.