It is the administration task we love to hate: securing a website. Apache forms the backbone of most websites so it makes sense to start there. In Hardening Apache, Tony Mobily does just that, starting with the basics of creating of a secure Apache installation and moving on to more in depth techniques for securing Apache installations from attack. Let’s see what Tony has to say when I talk to him about his new book and how to approach security, Apache and otherwise.

One of the key elements I get from your book is the back to basics approach. For example, I know a lot of companies with extensive login systems that leave their server room doors wide open. Do you it’s best to work from the inside out or the outside in when setting up security?

I believe that you always need to get the right person for the job. For example, if you need to re-tile your bathroom, you don’t call a wood worker. It’s the same with computer security; "physical" security (e.g. preventing people from breaking in) and "logical" security (preventing crackers and script kiddies from using your servers and resources) are very different things which require very different skills and training.

In this field - in fact, in any field - improvisation is just not an option.

If a company asked me to secure their physical network, I would redirect them to Steve, a friend of mine who does just that. Steve tells me amazing stories of sniffing packets by just placing a device next to the cable, for example, or other stories which I would see nicely in a James Bond movie rather than real life.

Even "logical" security branches out! I wouldn’t be able to audit the source code of a complex program, for example.
The problem is that even though improvisation shouldn’t be an option, it still happens. When a manager installs updates on a Unix system, or (worse) a service pack on a Windows machine, he is improvising and putting his systems at risk - full stop.

To go back to the question, security is a problem that has to be faced as a whole. To connect to the example I made earlier, a good physical design will prevent problems such as random people getting to close to a network cable and sniffing packets, or people accessing the servers’ consoles. On the other hand, a good logical design will mean that any piece of information will be encrypted, and if intruders did manage to access the cable, they won’t be able to do anything with the collected information.

Apache 1.3.x or Apache 2.x?

For me, there is no doubt: Apache 2.x.
It’s not just a matter of wanting to use the latest piece of software at any cost.

The problem with security is that often you are tied to the Apache version you are using. For example, if you use Apache 1.3.x for long enough on a complex web site, eventually you will be using a number of modules which are only available on Apache 1.3.x. In this common situation, upgrading to Apache 2.x can be really hard and might even require redesigning some parts of your web sites in order to use different technologies. The longer you leave it, the harder it will be to actually upgrade.

The problem is that eventually, you will have to upgrade because the 1.3.x branch of Apache will no longer be supported and patched anymore. It might not be soon, but it will happen. A lazy system administrator, at that point, will find himself (or herself) with an unpatchable system and, what’s worse, he or she won’t be able to upgrade without majorly disrupting the hosted web sites.

You make good use of the warnings and notifications made by sites like CVE and ApacheWeek. Are these sites that Apache administrators should be checking regularly?

Yes, absolutely.
Checking sites like ApacheWeek is both necessary and boring. I think there is also fear - sometimes you are just about to go on holiday or go home, and you discover that your production server has a security hole as big as a crater, and you urgently need to recompile the whole thing!
These sites are crucial to make sure that system administrators don’t live in their own "little world", and can realize that software is not just something they install on their computers and it works; software changes, evolves, improves, stumble across problems, and so on.

You use a lot of sample exploits to demonstrate weaknesses. Is it worth creating a tool-kit for checking these exploits against your site?

Writing such a tool-kit is a good idea in theory. In practice, however, there isn’t really much point because you know that if you upgrade your Apache server when you need to, then the security problems will be fixed.

What would you say was the weakest part, security wise, of most websites ?

That’s a hard question! It took me a while to work out what the most sensible answer is: the weakest part is the lack of maintenance and upgrade.
The problem is that keeping a system updated is hard work. If you manage 40, 50, or 150 Unix systems, then keeping up with all of them does require a whole lot of skills, because at that point the shell is just not good enough. You need to use something like CFEngine to configure them, and other automated tools to keep an eye on their security.

Here is an example: I have my own server, where I host my personal web site, my friends’ email, their small sites and so on.
I receive my email from LogWatch every day.

Today, it read:

**Unmatched Entries**
Illegal user patrick from 161.53.202.3
Illegal user patrick from 161.53.202.3
Illegal user rolo from 161.53.202.3
[…]
Illegal user john from 161.53.202.3
Illegal user test from 161.53.202.3
Illegal user merc from 151.31.36.81

Normally, I would run whois, find out who manages those networks, and report these attempts. Well, today I simply didn’t have the time. I am writing this answer on a train to London. Tomorrow I will be in Brunei, and in three days I will be back to Perth. My Internet connection is expensive and erratic. So here I am, Mr. 161.53.202.3 tried to attack me and he won’t be reported. And that’s only one person (me) with only one server!

Do you advocate the use of ‘cracker’ tools for testing?

That’s another tough one.

Well, I don’t advocate the /use/ of such tools. However, I am strongly against making these tools illegal.

Crackers and script kiddies, at the end of the day, are our friends (!). If you compared the Internet to an living organism, they are like those nasty (biological) viruses which occasionally knock you down and give you a tremendous sore throat, but are necessary to keep your body alert and your antibodies "trained". Also, if you catch a cold you can’t blame it completely on the virus - you’ve got to wonder if your body is healthy.

I believe it’s the same with the Internet: crackers will randomly try and get into your system (literally!). You have to make sure your defences are strong enough and well organized, so that when that happens you are prepared.

Some big companies won’t accept that. They will try to make tools such as Nexus illegal. Why? Maybe because they think that if such tools are not available anymore, then crackers will simply disappear. Or who knows, maybe they would like to sell testing tools to certified companies for a lot of money…

You cover quite a few security modules. Which would pick, and why, as the best modules?

The best and most useful module in my opinion is mod_dosevasive written by Jonathan A. Zdziarski. I believe Jonathan deserves a monument dedicated to him, also because he wrote DSPAM (which saves my life on a daily basis).

I believe that it should be part of the default Apache installation - in fact, I wonder if the Apache group would.

Tell us what your ‘Apache in Jail’ chapter is all about.

Well, jailing can be extremely complicated, but at the same time it is a very powerful tool against crackers.

Thanks to the system call "chroot()", you can tell a program what the root directory is when it runs. For example, you could run Apache making it believe that the root directory ("/") is "/cage/apache". This means that Apache will not be able to see anything outside "/cage/apache" - which is while you say that it’s "jailed". If a cracker does manage to use a buffer overflow exploit against your server, and get Apache to execute arbitrary commands, there will be nothing in /bin or /sbin to be executed, because /cage/apache/bin and /cage/apache/sbin will be nearly empty!

In my book, I tried to explain how to "jail" Apache step by step, by trying to make the readers aware of why and how everything was done. This deep understanding is necessary, because it is really quite tricky to use more complex software and third party modules on a jailed Apache.

You have some unusual outside interests. How did you end up sharing your life between Apache security with Jazz and Ballet?

At the moment it looks like I have a broken knee and I haven’t danced in ages (2 months), which is very sad. Classical Ballet (I am hopeless at jazz) has become part of me. As you can imagine, I spend a lot of time sitting down in front of a computer. Dancing is my escape: I love classical music, and I love feeling fit. You see, when you are training at ballet, you are a sort of a super-human: you never get tired, you are very flexible, and you generally feel good (and as this is a serious interview, I won’t mention the fun of making six pirouettes suddenly in the middle of the footpath while having a stroll with friends).

It’s funny, because I never considered the two things (computers and dancing) to be in contrast.

Anything in the pipeline?

Well, right now I am working on "Free Software Magazine" (www.freesoftwaremagazine.com), a magazine which concentrates entirely on free software.
It has been amazingly challenging. The first issue (January 2005) required a huge effort from many people, but the result is really rewarding.

Tony Mobily Bio

Tony Mobily is the project coordinator of Free Software Magazine.

When he is not talking about himself in the third person, Tony Mobily, BSc, is an ordinary human being, enjoying his life in the best city in the world: Perth (Western Australia). He is a senior system administrator and security expert, and is knowledgeable in several internet technologies. He loves Linux, Apache, Perl, C, and Bash.

Tony has been in the publishing industry his whole life, starting from the Italian magazine Dev. (he is lucky enough to be bilingual) in 1996.

He is also trained in Classical Ballet (ISTD), and fighting his way through learning hip hop and jazz. He also writes short and long stories, and keeps a blog at http://www.mobily.com.

Install Linux and the chances are you’ll be given the choice between a GNOME or KDE desktop. GNOME is the better known of the two, but if you want to development applications that use the GNOME environment where do you start? Well a good place would be Matthias Warkus’ new book, The Official GNOME 2 Developers Guide. I talk to Matthias and ask him about the GNOME system and environment, along with one or two other topics.

Could you describe to us what GNOME is?

GNOME is one of the leading projects developing user-friendly free software. The GNOME community effort includes the GNOME Desktop & Developer Platform, probably the most advanced free desktop environment around, translations, documentation and many third-party applications.

What you actually see on a computer said to be "running GNOME" is a tightly integrated, no-frills desktop system, on par with any commercial offering.

What is the benefit of the GNOME system over more traditional window managers, like Motif?

Actually, neither GNOME nor Motif are window managers, though both include one

The difference is so huge it’s hard to decide where to start. Not only is GNOME’s basic GUI technology (GTK+) much more advanced than the Motif toolkit (it can, for example, display right-to-left scripts such as Hebrew or CJK scripts such as Chinese), but the overall goal of the system is much more ambitious. What GNOME is trying to do is to integrate all system components well, and not in the traditional Unix way of providing a default that will work in 90% of all cases, whereas in all other cases, something has to be fixed by hand; GNOME intends to completely and "Just Work" in all supported environments.

You can witness this sort of integration in the new GNOME support for removable media. Whatever you insert or plug into the system, be it a CD, DVD, digital camera or USB stick, it will instantly be recognised and an appropriate window to access it will be opened.

GNOME seems to encompass a lot more than window dressing. Using your book I was able to create quite complex applications with some fairly advanced widgets with less than a hundred lines of code. Is this fairly common of the GNOME environment?

The GTK+ library stack, which sits at the core of GNOME, includes very powerful widgets, such as the file chooser, colour picker etc., but especially the text and tree/column view widgets based on the model-view-controller paradigm.

Other GNOME libraries bring even higher-level functionality. GNOME tries as much as possible to prevent programmers from reinventing the wheel.

You’ve managed to get a good balance in the book between the examples and the reference material. Do you have a favourite example from the book?

I suppose my favourite example would be the GdkPixbuf demo (pp. 132-136), a little thingy that lets you can change the scale and saturation of an image with two sliders. I think it’s less than 300 lines, half of which is comments and whitespace. Very neat example, and impressively small in size considering it’s pure C and not something higher-level such as Python or Java.

Could you tell us a bit more about gconf?

GConf is the solution to the recurring problem of where to store and how to sensibly process configuration values. It’s a self-documenting, typesafe database with a tree structure that is usually saved in XML. Applications connect handlers to GConf keys, and any changes to a key, whether from the app itself, another instance of it or an external configuration tool, will at once apply to all running instances. There is a GConf editor to centrally change all system settings. Default settings can be provided and mandatory settings can be enforced centrally, for all users. The new GConf editor includes special administrator functions to do this. This is essential for the large installations where GNOME is getting popular these days: There are organisations rolling out GNOME on several tens of thousands of desktop computers.

I suppose you could call GConf the Windows registry done right. People used to hate central configuration databases because the nightmare that is the Windows registry was the only one they knew. GConf is starting to change that. It’s really a good idea.

I’ll admit to being new to GnomeVFS. Is this something that could be adopted wider amongst the Linux community?

Actually, because GNOME is not Linux-only, it would need to be adopted across a broader set of platforms. Perhaps the people working on cross-desktop standards specifications at freedesktop.org will make this real one day, who knows?

Anyway, GNOME-VFS is a very nice interface to access files in a network-transparent and asynchronous way. I think it’s performance has improved a lot over the last months, too. Writing a GNOME application, there’s no reason to use the old libc file access functions anymore; using GNOME-VFS, your application will, at no extra cost, be able to process remote files as well as local files.

Your book focuses on the C API. Are there any other alternatives?

GNOME has officially supported language bindings for C++ (making use of all C++ features in the canonical way, unlike, for example, Qt), for Java, Perl and Python. Especially Python is popular as an RAD language; in combination with the Glade user interface builder, you can write productive GNOME applications in no time flat.

Unofficial language bindings exist for many languages, including exotic ones; there are bindings for (at least) C#, D, Eiffel, Erlang, Euphoria, Felix, Gauche, Guile, Haskell, JavaScript, Objective-Caml, Pascal, Pike, PHP, Ruby, Scheme, S-Lang, Smalltalk, Tcl, TOM and XBase, though the degree of support varies widely.

Programming GNOME in C# is becoming popular, and the Ruby bindings do also seem to have some success.

What do you think of KDE?

GNOME would not exist without KDE. Probably free software’s desktop ambitions wouldn’t be as visible as they are today. We all owe a great deal to the KDE project.

Being used to GNOME, most KDE applications look confusing to me. I like GNOME’s philosophy of keeping user interfaces as lean as possible. One example: In the default setting, KDE’s file manager presents so many toolbar, sidebar and status bar icons to me that I instinctively want it to just go away again.

I also like GNOME’s way of keeping the number of distinct user-visible components low by integrating new functionality into existing applications. Work is currently being done on a CD/DVD burning framework that will integrate audio CD burning into the audio player etc.; unlike KDE, we don’t think writing an all-singing, all-dancing Nero clone with four different configuration dialogues, several toolbars and theme-able icons is the way to go. Don’t get me wrong - I seriously love the job they did on the underlying functionality and I use K3B all the time. But I don’t really think the interface is appropriate.

Is there a future for both alternatives, or do you seem some kind of merging in the future?

GNOMEs well-integrated, no-nonsense desktop with the excellent Evolution groupware client and several administration and lockdown features seems to be better suited to the large-scale free software desktop deployments we are seeing at the moment than KDE.

I don’t think that KDE will ever go away. Neither will GNOME, for that matter. Some years ago, many KDEers kept telling GNOME to just fold and merge with KDE; I don’t hear this anymore. With different views on how user interfaces should look and work and how functionality should be distributed across the system, there is a place in the world for both.

What’s your favourite cartoon character?

Hard question. I suppose that would be Piro from MegaTokyo, with Warren from Absurd Notions and Cerulean the Dragon from "Why the long face" a close second and third.

What are you working on at the moment?

I’m not really working on any GNOME-related things at the moment. My focus is on getting on with my studies; I’m in my fifth semester of philosophy, sociology and French, and my main activity is to learn ancient Greek, which is taking up much of my spare time.

I intend to review the original German edition of my book for an eventual revised and extended second edition, but it seems it’ll be hard to find a publisher, and I haven’t got the time at the moment anyway. I hope someday I’ll have more time to consecrate to working and writing on GNOME.

Matthias Warkus Bio

Matthias Warkus was born and raised in one of the most rural regions of Germany. He started using Linux out of sheer boredom at the age of sixteen. Shortly afterwards, he got involved with GNOME, first as a translator, later also doing promotional work, holding many GNOME-related talks in Germany. He considers himself to be better at writing than at coding, and thus went on to write the Official GNOME 2 Developer’s Manual. Currently, he is a student of philosophy. When he’s not struggling with lofty theories in class or discussing them with his friends in one of Marburg’s countless pubs, he enjoys reading, writing and playing the piano.

Knoppix is not just another Linux distribution. Unlike many Linux alternatives, Knoppix doesn’t need to be installed; everything runs from a CD (called a ‘Live CD’ distribution). While Live CDs aren’t unique to Knoppix, it is the way the Knoppix CD is packaged that makes the difference. Knoppix includes intelligent hardware detection – it can automatically identify nearly everything on your machine and then make the bet of it – and the CD includes a wide selection of programs, from typical Linux applications through to repair utilities and tools.
In Knoppix Hacks you get 100 tips on using Knoppix, from simply running Knoppix through to customizing your Knoppix installation, repairing Linux and Windows machines using Knoppix and setting up emergency, Knoppix based, routers, file and web servers.
In the same way that Knoppix itself is immensely useful in hundreds of different ways, so is this book.

Knoppix Hacks packs a lot of information into a very small space; the book isn’t quite pocket sized, but it’s the smaller format paperback (slightly wider than the Knoppix CD included inside the back cover) and just 300 pages long. Don’t let this fool you though. The book is organized a bit like an FAQ – although organized for tips, rather than questions and answers. They range from a simple two page guides to running Knoppix up to more extensive multi-page tips on specific topics. The book is grouped into sections, starting with the simple tasks of booting Knoppix and using it as a desktop operating system. We then move onto the real meat of the book; using Knoppix in an emergency. Two final sections then look at more advanced techniques and customizing your own Knoppix CD with your favourite list of software and tools.

The book (and Knoppix) will appeal to a wide range of people. Linux users will appreciate the speed with which you can be running within a Linux environment, on pretty much any machine, using the enclosed CD.
If you use Knoppix as your desktop operating system (and many do, because of it’s wide support), then the tips in the book will be useful, but you shouldn’t use it as a guide to using Linux. There are tips for desktop users, for example for multimedia playback, browsing the Internet (including a useful guide to mobile GPRS users), but the main focus is definitely at the more expert user and system administrator.
Because Knoppix runs entirely from the CD it makes an ideal solution when you need an emergency system, but don’t have time to install it, or when you want to repair an existing system, and can’t do so using the existing operating system tools. Knoppix Hacks shows you how to use Knoppix to set up a quick server, for example for file sharing or web serving. It also includes extensive details on using Knoppix when repairing both Linux and Windows systems.
There’s a good gamut of topics here. For example, under Linux you can repair Lilo and Grub boot loaders, repair and reconfigure partitions and filesystems and even tips on migrating to new hard drives or RAID volumes. For Windows users there are tips for repairing the bootloader, re-organize partitions and even tools and information on editing the Windows registry, NT passwords and downloading patches. Put together, the repair tools alone should get you out of the majority of holes with relative ease.

In the same way that Knoppix itself is immensely useful in hundreds of different ways, so is this book. It is hard not to get hooked on trying different things from the book, even if you don’t really have a direct need for them.
The books content and style is easy to use and understand and each hack is both straightforward and detailed. There’s more than just a quick overview of what needs to be done here; you get side information, background details and dependency information too. If a particular hack needs more than Knoppix (for example, the wardriving example also needs a GPS system and USB to serial adaptor), then information is included on that too.

If there is a downside to the book it’s that it ends just slightly to quickly. Once you start using Knoppix and the tips in the book you want to know how to do more and more. The books tone is light and welcoming, and so packed with information that it comes as quite a surprise when book is over.
This is far from a criticism, but I’m certainly hoping for a ‘Knoppix Hacks Volume 2’ in the near future.