Secure Coding: Principles and Practices [Paperback]

In the same way, Mark G. Graff and Kenneth R. van Wyk have provided an excellent book that gives us a framework for thinking about security rather than trying to give specific rules that might have been invalid before the book came off the press. "Secure Coding" gives the reader the ability to envision, architect, design, code, and implement a security framework that truly meets the needs of its stakeholders.

The authors don't provide a cookbook. In their own words: "When you picked up this book, perhaps you thought that we could provide certain security? Sadly, no one can."

Instead, they deliver a robust mental model and a framework to understand security and to architect, design, develop, and operate secure systems. They present best practices in the field of security, the reasons for using them, and suggestions on deciding which practices are appropriate in your particular case.

Their approach is to realize that the objective is not to make a system totally secure, but to make it just secure enough. Deciding what is "just secure enough" is a business and not a technical decision. It is based on weighing risk versus cost.

There are substantial references throughout the book as well as an appendix of resources. The book is filled with examples of security failures and, more importantly, an excellent post mortem on each to show what could have been done to avoid the problem. The authors are extremely familiar with UNIX environments and this comes through in the examples. However, you don't need to be a UNIX guru to glean valuable lessons from the examples.

One key message is that security is not something you can bolt onto an application. You must take a holistic approach to the overall system in which the application is being used. It's worth noting that many secure applications become extremely insecure because of the system environment (including networks) in which they exist.

A second key message is that, while you can retrofit a insecure application, it is far easier and far less costly to incorporate security as an integral part of the entire development life-cycle including requirements, architecture, and design. The security architecture and design must be well-documented so that future maintenance does not inadvertently introduce gaping security holes.

The book is primarily intended for those who architect, design, and code secure applications. However, I believe that it is a must read for those who manage and those who implement secure applications and systems.

Passwords, encryption, and all the rest are the lock. This book is more about making the door and frame strong. Remember the Blaster worm? That wasn't a 'security' problem. It exploited bugs in Windows that supposedly had nothing to do with security.

This book is about building programs that resist attack. That doesn't mean copying a safe code fragment into your program and declaring it safe - that idea is ludicrous. Instead, this book is about the process that designs and implements strong programs. It starts with architecture and design documents, then follows through to design and maintenance.

The weakness of this book is lack of detail - how to build fail-safe code, what needs to be on design and inspection checklists, etc. There's good reason for that: each sub-topic needs books, if not whole libraries of its own. Take fault tolerance, for example. It may not sound like security, but an attack is meant to cause system failures, and fault tolerance is design to withstand failures. Fault tolerance is a huge topic, with journals and literature all its own. This book can barely mention the idea, while still giving other topics their due. It's a start, though.

Much of the advice may sound drearily familiar: code reviews, security audits, configuration control, error checking, and all the other things that take the 'fun' out of programming. If people want that kind of 'fun', then stop calling them software engineers. They're not ready for adult responsibilities.

Before anything else, software security requires correct behavior from a program. I really hope I don't hear objections to that as a basic design goal.

This book teaches you how to *think* about security, how to think about and *design* code that will be secure. It isn't a "add this snippit of code to your input buffer validation function" sort of book. There are many of these books, and they're useful in their place, but this book writes about the design of secure code, not the actual specifics.

To continue the cooking analogy, this is a book on how to write receipes, not a book *of* receipes.

Disclaimer, I helped review this book - and I think it's the sort of work that has been sorely missing in the field (I was also given a free copy for doing the review work).

Jeremy Allison,
Samba Team.