Code Review Introduction
Code review is probably the single-most effective technique for identifying security flaws. When used together with automated tools and manual penetration testing, code review can significantly increase the cost effectiveness of an application security verification effort.
This document does not prescribe a process for performing a security code review. Rather, this guide focuses on the mechanics of reviewing code for certain vulnerabilities, and provides limited guidance on how the effort should be structured and executed. OWASP intends to develop a more detailed process in a future version of this guide.
Manual security code review provides insight into the “real risk” associated with insecure code. This is the single most important value from a manual approach. A human reviewer can understand the context for certain coding practices, and make a serious risk estimate that accounts for both the likelihood of attack and the business impact of a breach.
Why Does Code Have Vulnerabilities?
MITRE has catalogued almost 700 different kinds of software weaknesses in their CWE project. These are all different ways that software developers can make mistakes that lead to insecurity. Every one of these weaknesses is subtle and many are seriously tricky. Software developers are not taught about these weaknesses in school and most do not receive any training on the job about these problems.
These problems have become so important in recent years because we continue to increase connectivity and to add technologies and protocols at a shocking rate. Our ability to invent technology has seriously outstripped our ability to secure it. Many of the technologies in use today simply have not received any security scrutiny.
There are many reasons why businesses are not spending the appropriate amount of time on security. Ultimately, these reasons stem from an underlying problem in the software market. Because software is essentially a black-box, it is extremely difficult to tell the difference between good code and insecure code. Without this visibility, buyers won’t pay more for secure code, and vendors would be foolish to spend extra effort to produce secure code.
One goal for this project is to help software buyers gain visibility into the security of software and start to effect change in the software market.
Nevertheless, we still frequently get pushback when we advocate for security code review. Here are some of the (unjustified) excuses that we hear for not putting more effort into security:
“We never get hacked (that I know of), we don’t need security”
“We have a firewall that protects our applications”
"We trust our employees not to attack our applications"
Over the last 10 years, the team involved with the OWASP Code Review Project have performed thousands of application reviews, and found that every single application has had serious vulnerabilities. If you haven’t reviewed your code for security holes the likelihood that your application has problems is virtually 100%.
Still, there are many organizations that choose not to know about the security of their code. To them, we offer Rumsfeld’s cryptic explanation of what we actually know. If you’re making informed decisions to take risk in your enterprise, we fully support you. However, if you don’t even know what risks you are taking, you are being irresponsible both to your shareholders and your customers.
"...we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns -- the ones we don't know we don't know." - Donald Rumsfeld
What is Security Code Review?
Security code review is the process of auditing the source code for an application to verify that the proper security controls are present, that they work as intended, and that they have been invoked in all the right places. Code review is a way of ensuring that the application has been developed so as to be “self-defending” in its given environment.
Security code review is a method of assuring secure application developers are following secure development techniques. A general rule of thumb is that a penetration test should not discover any additional application vulnerabilities relating to the developed code after the application has undergone a proper security code review.
All security code reviews are a combination of human effort and technology support. At one end of the spectrum is an inexperienced person with a text editor. At the other end of the scale is a security expert with an advanced static analysis tool. Unfortunately, it takes a fairly serious level of expertise to use the current application security tools effectively.
Tools can be used to perform this task but they always need human verification. Tools do not understand context, which is the keystone of security code review. Tools are good at assessing large amounts of code and pointing out possible issues but a person needs to verify every single result to determine if it is a real issue, if it is actually exploitable, and calculate the risk to the enterprise.
Human reviewers are also necessary to fill in for the significant blind spots where automated tools simply cannot check.
Laying the ground work
In order to effectively review a code baseline, it is critical that the review team understands the business purpose of the application and the most critical business impacts. This will guide them in their search for serious vulnerabilities. The team should also identify the different threat agents, their motivation, and how they could potentially attack the application.
All this information can be assembled into a high-level threat model of the application that represents all of information that is relevant to application security. The goal for the reviewer is to verify that the key risks have been properly addressed by security controls that work properly and are used in all the right places.
Ideally the reviewer should be involved in the design phase of the application, but this is almost never the case. More likely, the review team will be presented with a large codebase, say 450,000 lines of code, and will need to get organized and make the best possible use of the time available.
Performing code review can feel like an audit, and most developers hate being audited. The way to approach this is to create an atmosphere of collaboration between the reviewer, the development team, the business representatives, and any other vested interests. Portraying the image of an advisor and not a policeman is very important if you wish to get full co-operation from the development team.
Security code review teams that successfully build trust with the development team can become trusted advisor. In many cases, this will lead to getting security folks involved earlier in the lifecycle, and can significantly cut down on security costs.
Before we start
The reviewer(s) need to be familiar with:
- Code: The language(s) used, the features and issues of that language from a security perspective. The issues one needs to look out for and best practices from a security and performance perspective.
- Context: The working of the application being reviewed. All security is in context of what we are trying to secure. Recommending military standard security mechanisms on an application that vends apples would be over-kill, and out of context. What type of data is being manipulated or processed and what would the damage to the company be if this data was compromised. Context is the "Holy Grail" of secure code inspection and risk assessment…we’ll see more later.
- Audience: The intended users of the application, is it externally facing or internal to “trusted” users. Does this application talk to other entities (machines/services)? Do humans use this application?
- Importance: The availability of the application is also important. Shall the enterprise be affected in any great way if the application is “bounced” or shut down for a significant or insignificant amount of time?
Discovery: Gathering the information
The review team will need certain information about the application in order to be effective. The information should be assembled into a threat model that can be used to prioritize the review. Frequently, this information can be obtained by studying design documents, business requirements, functional specifications, test results, and the like. However, in most real-world projects, the documentation is significantly out of date and almost never has appropriate security information.
Therefore, one of the most effective ways to get started, and arguably the most accurate, is to talk with the developers and the lead architect for the application. This does not have to be a long meeting, but just enough for the development team to share some basic information about the key security considerations and controls. A walkthrough of the actual running application is very helpful, to give the review team a good idea about how the application is intended to work. Also, a brief overview of the structure of the codebase and any libraries used can help the review team get started.
If the information about the application cannot be gained in any other way, then the team will have to spend some time doing reconnaissance and sharing information about how the application appears to work by examining the code.
Context, Context, Context
Security code review is not simply about reviewing code. It’s important to remember that the reason that we review code is to ensure that the code adequately protects the information and assets it has been entrusted with, such as money, intellectual property, trade secrets, lives, or data.
The context in which the application is intended to operate is a very important issue in establishing potential risk. If reviewers do not understand the business context, they will not be able to find the most important risks and may focus on issues that are inconsequential to the business.
As preparation for a security code review, a high level threat model should be prepared which includes the relevant information. This process is described more fully in a later section, but the major areas are listed here:
- Threat Agents
- Attack Surface (including any public and backend interfaces)
- Possible Attacks
- Required Security Controls (both to stop likely attacks and to meet corporate policy)
- Potential Technical Impacts
- Important Business Impacts
Defining context should provide us with the following information:
- Establish the importance of the application to the enterprise.
- Establish the boundaries of the application context.
- Establish the trust relationships between entities.
- Establish potential threats and possible controls.
The review team can use simple questions like the following to gather this information from the development team:
“What type/how sensitive is the data/asset contained in the application?”:
This is a keystone to security and assessing possible risk to the application. How desirable is the information? What effect would it have on the enterprise if the information were compromised in any way?
“Is the application internal or external facing?”, “Who uses the application; are they trusted users?”
This is a bit of a false sense of security as attacks take place by internal/trusted users more often than is acknowledged. It does give us context that the application should be limited to a finite number of identified users, but it’s not a guarantee that these users shall all behave properly.
“Where does the application host sit?”
Users should not be allowed past the DMZ into the LAN without being authenticated. Internal users also need to be authenticated. No authentication = no accountability and a weak audit trail.
If there are internal and external users, what are the differences from a security standpoint? How do we identify one from another? How does authorization work?
“How important is this application to the enterprise?”.
Is the application of minor significance or a Tier A / Mission critical application, without which the enterprise would fail? Any good web application development policy would have additional requirements for different applications of differing importance to the enterprise. It would be the analyst’s job to ensure the policy was followed from a code perspective also.
A useful approach is to present the team with a checklist, which asks the relevant questions pertaining to any web application.
Defining a generic checklist which can be filled out by the development team is of high value, if the checklist asks the correct questions in order to give us context. The checklist is a good barometer for the level of security the developers have attempted or thought of. The checklist should cover the most critical security controls and vulnerability areas such as:
- Data Validation
- Session Management
- Error Handling
- Security Configuration
- Network Architecture