Client-side MYSQL queries - C++ - bad idea?

Started by
6 comments, last by Nypyren 11 years, 1 month ago

Hello everyone!

I was wondering how much of a security risk it would be to connect to a MySQL database via client-side only.

I could make a client/server app but I don't really have the funds to get a server at this time.

Thank you for your input in advance.

Tyler

Advertisement

If you don't have a server, then it doesn't matter. The security risk is when you have a client, a server, and allow the client to send arbitrary SQL commands to the server. A malicious user notices this and then decides to just send whatever SQL command he wants, such as "select * from creditcards" (or whatever - you get the idea).

If you only have a client, and no server, the malicious user already has access to everything he could want, so whether you use SQL or not won't matter in that regard.

If you don't have a server, then it doesn't matter. The security risk is when you have a client, a server, and allow the client to send arbitrary SQL commands to the server. A malicious user notices this and then decides to just send whatever SQL command he wants, such as "select * from creditcards" (or whatever - you get the idea).

If you only have a client, and no server, the malicious user already has access to everything he could want, so whether you use SQL or not won't matter in that regard.

I was originally going to have the client send information/variables to the server and only have the server do queries.

I guess I'll have to stop buying my kids toys for a little :P

I was originally going to have the client send information/variables to the server and only have the server do queries.

Yeah. Only the server should perform actual queries.

Also, make sure that your server examines the variables the client sends to make sure they won't break anything. For example, if your software builds a query string, and inserts a variable that the client provided, make sure that the variable doesn't contain something that would make the query do something bad.

See this: http://en.wikipedia.org/wiki/SQL_injection
And this: http://www.unixwiz.net/techtips/sql-injection.html

There is one rule you should be aware of from the get-go that will save you a lot of pain (and potentially legal liability in the worst case):

Nothing the client ever sends you should ever be trusted, full stop. No argument, no exceptions.

There is also an important corollary to this that you should keep in mind:

Everything the client can see is fair game to someone who wants to abuse your system.

The more visibility the client has into how things work, the more likely it is that someone can do something they shouldn't. (I am not advocating "security by obscurity" here - just pointing out that giving control to the client is a liability.)

By the same token, the more responsibility the client has, the more likely it is that someone can do something they shouldn't. You should always aim to design these sorts of things to where the client has a minimum of information and a minimum of responsibility.

SQL injection is a classic example of what can go wrong if you forget these two rules. It's just the beginning, though, and defending against it is actually notoriously hard to get right. (I actually recommend using parameterized SQL and ideally something like stored procedures, instead of constructing SQL from strings, because of this.)

Wielder of the Sacred Wands
[Work - ArenaNet] [Epoch Language] [Scribblings]

MySQL is not a hardened system in the sense that it is guaranteed safe against abuse when exposed to the greater internet. You should not let clients talk directly to it -- always put an application server of some sort (game server, web server, ...) in the middle to verify requests and shield your database from the hostile place that's the internet!
enum Bool { True, False, FileNotFound };

I think it's really a bad idea. Even if you give them access to no tables - only stored procedures - they can STILL break the server. There are numerous well-documented ways of either crashing the server or making it consume all its resources.

The INFORMATION_SCHEMA tables are visible to everyone regardless of permissions, and they can still do SELECTs on them. So if they join one of the (more than one row) I_S tables with itself a dozen times, then the server will attempt to create a result set bigger than your disc (in temp storage normally).

I think it's really a bad idea. Even if you give them access to no tables - only stored procedures - they can STILL break the server. There are numerous well-documented ways of either crashing the server or making it consume all its resources.

The INFORMATION_SCHEMA tables are visible to everyone regardless of permissions, and they can still do SELECTs on them. So if they join one of the (more than one row) I_S tables with itself a dozen times, then the server will attempt to create a result set bigger than your disc (in temp storage normally).

I think ApochPiQ meant that he only allows the client to talk to the server service (not the SQL database itself) AND the server's service only runs stored procedures with sanitized parameters instead of raw SQL queries.

This topic is closed to new replies.

Advertisement