Finding Vulnerabilities in Software

I'm insterested to know the techniques that where used to discover vulnerabilities. I know the theory about buffer overflows, format string exploits, ecc, I also wrote some of them. But I still don't realize how to find a vulnerability in an efficient way.

I don't looking for a magic wand, I'm only looking for the most common techniques about it, I think that looking the whole source is an epic work for some project admitting that you have access to the source. Trying to fuzz on the input manually isn't so comfortable too. So I'm wondering about some tool that helps.

E.g.

I'm not realizing how the dev team can find vulnerabilities to jailbreak iPhones so fast. They don't have source code, they can't execute programs and since there is a small number of default programs, I don't expect a large numbers o开发者_如何转开发f security holes. So how to find this kind of vulnerability so quickly?

Thank you in advance.

On the lower layers, manually examining memory can be very revealing. You can certainly view memory with a tool like Visual Studio, and I would imagine that someone has even written a tool to crudely reconstruct an application based on the instructions it executes and the data structures it places into memory.

On the web, I have found many sequence-related exploits by simply reversing the order in which an operation occurs (for example, an online transaction). Because the server is stateful but the client is stateless, you can rapidly exploit a poorly-designed process by emulating a different sequence.

As to the speed of discovery: I think quantity often trumps brilliance...put a piece of software, even a good one, in the hands of a million bored/curious/motivated people, and vulnerabilities are bound to be discovered. There is a tremendous rush to get products out the door.

There is no efficient way to do this, as firms spend a good deal of money to produce and maintain secure software. Ideally, their work in securing software does not start with a looking for vulnerabilities in the finished product; so many vulns have already been eradicated when the software is out.

Back to your question: it will depend on what you have (working binaries, complete/partial source code, etc). On the other hand, it is not finding ANY vulnerability but those that count (e.g., those that the client of the audit, or the software owner). Right?

This will help you understand the inputs and functions you need to worry about. Once you localized these, you may already have a feeling of the software's quality: if it isn't very good, then probably fuzzing will find you some bugs. Else, you need to start understanding these functions and how the input is used within the code to understand whether the code can be subverted in any way.

Some experience will help you weight how much effort to put at each task and when to push further. For example, if you see some bad practices being used, then delve deeper. If you see crypto being implemented from scratch, delve deeper. Etc

Aside from buffer overflow and format string exploits, you may want to read a bit on code injection. (a lot of what you'll come across will be web/DB related, but dig deeper) AFAIK this was a huge force in jailbreaking the iThingies. Saurik's mobile substrate allow(s) (-ed?) you to load 3rd party .dylibs, and call any code contained in those.