EncFSVault provides a replacement for Apple’s FileVault. There are a lot of issues with FileVault. Personally I don’t like any proprietary software for security sensitive storage of my data. But the main reason I was not able to use FileVault was the fact that FileVault still doesn’t provide support for case sensitive HFS+ file systems as of OS X Leopard 10.5.6. That’s a shame!
My choice was EncFSVault. Good or bad choice?

Read the rest of this entry »

In this tutorial I’d like to describe how to create an extensive sendmail wrapper for a web server to monitor all sent emails and throttle daily sent email volume by the senders original UID (user id). This is useful if you e.g. run PHP in CGI-mode with SuExec, that is: all customers run their scripts under their own UID. The wrapper described here is not just a PHP-only wrapper (as described in my Simple PHP mail wrapper tutorial) – it directly replaces /usr/sbin/sendmail so we are able track all sent email of the whole system.

Read the rest of this entry »

I have updated the process hiding kernel patch described in my previous post. The following patch was created for a vanilla kernel 2.6.27.10 from kernel.org. You should be able to patch also future kernel versions as it is a super simple single-line patch. Just modify one line in the proc_pid_instantiate method of fs/proc/base.c.

Read the rest of this entry »

If you run a webserver with several hundreds of virtual hosts running PHP, you definitely need to monitor or log the access to PHP’s mail() function. I describe in a short tutorial how to painlessly setup a simple sendmail wrapper to accomplish this.
This has been tested on a Debian Lenny 5.0 system running PHP 5.2.8 and Postfix.

Read the rest of this entry »

Currently all Linux kernel security patch projects seem to be sleeping. There is no useful kernel patch that provides us with a decent patch set allowing us to strengthen the Linux kernel. Some years ago I was using Grsecurity, a wonderful solution to enforce security on 2.4.x kernels at that time. The project seems to be pretty dead by now.

During the last months I was using RSBAC, a great set of security enhancements to the 2.6.x kernels. RSBAC seems to be a great project and I like the way they provide pre-patched vanilla kernels. But again, reaction time is way too slow. Root exploits for Linux kernels seem to appear all the time and force a server administrator to react fast. The lately published vmsplice root exploit made me give up on RSBAC as it’s just always a step behind. I decided to switch back to self compiled vanilla kernels from kernel.org.

Read the rest of this entry »

A webserver usually is the primary target to intrude into any network. If you provide web hosting services for your customers you have to provide them with a lot of features to make them happy. The main requirement for any hosting provider is PHP, probably the widest spread web scripting language out there.

Some customers only start to get happy if you give them PHP without any safe_mode restrictions, if you provide them with custom CGI scripting next to the basic good old SSI features (which in my eyes no one really needs since we got PHP) by Apache HTTP Server, if you give them FTP access and let them manage their account by themselves.

In every feature there is always a hidden security risk. We cannot give all this to our customers without thinking about security and its consequences if a user gets hold of data which does not belong to himself or even breaks into the whole system. So, let’s start at the basics: No customer should be able to see any other running processes on the system except the ones that belong to himself. We want to hide all processes that the given customer is not allowed to see. That’s process hiding. And because on a Linux box it’s always smart to implement something from bottom up, we name it kernel based.

There is no simple solution for this problem. Some rootkits simply overwrite the ‘ps’ command. But we want something more trustworthy, somehow deeper anchored in the system (got that?). The only kernel patch I found was the one from RSBAC.org (Rule Set Based Access Control), a full blown kernel security patch. The only feature we actually need is “CAP process hiding”.

Read the rest of this entry »