Last week I happened to find a Barnes & Noble gift card in my wallet, with $75 left on it. What a wonderful discovery! One of the pile of books that I ordered was The Future of the Internet—And How to Stop It by Jonathan Zittrain. I’d fully intended to read the book cover-to-cover, perhaps even digest the content a little, before throwing my thoughts out there (presumably to get lost into the vast sea of crap that makes up the “blogosphere”). But I just have to get some thoughts down on paper…err…LCD.

First off, when I found out that Zittrain is a professor of Internet Law at Harvard, it immediately told me two things. First, that he probably sides with content producers and/or Big ‘Net a bit too much. Second, that he probably doesn’t really understand what the hell he’s talking about, or why people made the choices they have. The fact that the first chapter of the book, which talks about history, doesn’t seem to mention ARPAnet once only confirms this. But, the B&N summary sounded like the book has a healthy dash of iPhone bashing, so I figured it’s be a good read. It was also written in 2008, so I figured that the ideas would be relatively current.

Well, I’m just under a quarter of the way into the book, and given the vast mass of notes I’ve penned in the margins, I think Mr/ Zittrain and I wouldn’t get along too well on a desert island. But I’ll try to contain my commentary - and attacks upon the author - until I’m done with the book. The thought currently in my mind is a very specific one:

Many technologically savvy people think that bad code is simply a Microsoft Windows issue. They believe that the Windows OS and the Internet Explorer browser are particularly poorly designed, and that “better” counterparts (Linux and MacOS, or the Firefox and Opera browsers) can help protect a user. This is not much added protection. Not only do these alternative OSes and browsers have their own vulnerabilities, but the fundamental problem is that the point of a PC - regardless of its OS - is that its users can easily reconfigure it to run new software from anywhere.

To be sure, Microsoft Windows has been the target of malware infections for years, but this in part reflects Microsoft’s dominant market share.

Oh, wow, is it 2004 again? I thought we’d given up on the “market share” argument. When Apache had 10% of the market share of web servers, people said it wasn’t attacked as often because of low market share. Well, Apache currently has a 47% share of the market, compared to Microsoft’s 21%, and it’s still more secure, more stable, and has fewer critical vulnerabilities*. The same market share argument was made about Firefox when it had 5% market share. Now, the share is projected between 31.85% and 47%, and it still has less serious vulnerabilities (ones that can actually damage your computer) than Windows). I thought this “market share” argument was done with.

Most important is the thing that most Microsoft-biased pundits (and, of course, Microsoft themselves) don’t ever talk about: an amazingly large number of servers run Linux. Especially e-commerce servers which house loads of personal information and credit card numbers. Estimates for big e-commerce sites put non-Windows OSes at 30-50%, and they’re quite popular among small sites that probably don’t have well-trained SysAdmins. So, if Windows wasn’t really less secure, wouldn’t we see e-commerce servers getting compromised left and right?

But there’s a more important point here. It’s about curtailing the stupidity of users. I know, in Microsoft’s defense, that Windows Vista and Windows 7 are supposed to be better with this. But, at least in the past, Windows had virtually no privilege separation. With a little code, you could effect the whole system from an arbitrary binary - or worse, with ActiveX, through the browser. I was dumbfounded that any user could install a system-wide application. The real issue here, at least with older Windows (I don’t know much about the new ones) is that Windows, from the beginning, wasn’t written to be secure. Heck, it wasn’t even designed to be attached to a real network.

Linux does have real security advantages over Windows, and not just because it has low market share. First is an actual, true implementation of privilege separation. No matter what I do in my desktop web browser, no matter what I run, even if I installed a Firefox plugin that wanted to destroy my machine, it couldn’t happen. No matter what I let some random code do, it simply can’t escape the confines of my user account.

Ok, ok, I know what you’re all saying right now. I can hear it from here: “but what if the moron does everything as root? what if they just sudo anything that they’re asked about?” Well, I have answers to that, too. My own distro of choice, OpenSuSE greatly upset me when I went to install 11.1, and the installer showed a default of one user account, automatic login, and the same password for the user and root. That’s just stupid. In fact, it’s braindead, plain and simple. I don’t care how wonderful it would be to get Linux on every desktop in the world, if we have to destroy every advantage that Linux has over other OSes, it will be worthless.

I digress. In the end, it boils down to user education. And, in some respects, I think that Linux has become too dumbed-down. There are certain things that simply shouldn’t be put in a GUI. Excuse my elitism, but if you can’t figure out how to configure Apache correctly from the command line, you have no business running an Apache installation. The same goes for countless other services and applications. So, what’s my solution? Well, here’s what I do when I install Linux for non-technical friends. Some of these things are training items, others are things that I do in terms of configuration and, IMHO, should be OS/distro defaults (unless you know some esoteric hidden switch to change them).

  • Disable graphical login as root. This enforces proper use of sudo, and also prevents a user from becoming lazy and operating as root on a regular basis.
  • Pick a good, strong root password. Write it down on a post-it note and keep it somewhere near the computer. (Yes, I know what you’re thinking. But if it’s a home computer, anyone already in the house either is trusted, or will own the computer one way or another. I’d rather have everyone in the house have access to the box, than a password that a remote attacker can easily brute force.)*
  • Disable caching of sudo passwords in the desktop manager, if it already isn’t done. This is a very bad idea, IMHO, and effectively defeats privilege separation. If someone needs to use sudo that often, they’re either a knowledgeable user, or they’re doing something wrong.
  • Set the package manager to use the strictest key verification settings.
  • Provide the user with extensive documentation (can be a list of links to helpful sites) that includes - this is of paramount importance - a list of common Windows (or whatever OS they’re coming from) programs and their closest Linux equivalents. This is another measure to try and dissuade the user from searching for and installing arbitrary code.
  • Give the user a good, simple explanation of what sudo is, what root is, and why they should be worried. One of my analogies - if I have time to explain it - is to think of the computer’s security like a jewlery store. Your user account is the front door; only people who look honest are buzzed in, but they still can’t do much damage. The root password is the combination to the vault; only very trusted people can get in, and they only open it when they absolutely have to.
  • Enable a wide range of trusted repositories by default. The more likely the user is to find a package in the repos already cached, the less likely they are to download arbitrary code.
  • Explain to the user that when you install software (as root), you’re essentially giving the developer access to your system. Software should be screened by someone who knows what they’re doing (i.e. the community) before you install it.
  • I always tell people to *only* install software from the repositories I enable. If there’s something they need and it isn’t available, ask me (or ask the community) and I’ll make a package and upload it to a suitable repository. The key here - and the most difficult part - is to conquer the Windows habit of installing software from disparate sources, and train the user that only software from their repositories, or other community-standard repositories, can be trusted.
  • Show the user the correct patch/update procedure for their system. Depending on skill level and the level of attention you’re willing to give them, it might be advisable to enable automatic updates (if the OS doesn’t have a way to do it, then via cron).
  • If the user is a developer or needs to run any services, even just for development - i.e. Apache, MySQL, Postfix, etc. - properly secure them and give an overview and links to the proper security procedures.
  • Setup a second user account. Explain to the user that this is only to be used for banking and other sensitive activities. Lock it down, make sure it’s in a different group from the main user, don’t install any Firefox plugins.

Unfotunately, a lot of this is just breaking the bad administration and security habits shared by most Windows users.

While we’re on the topic, a word about package managers. I’m a Linux sysadmin, and I believe in ‘eating your own dog food’. I’ve used Linux on all of my servers, desktops, and laptops for over 4 years now. I haven’t used Windows on a regular basis in ages. I’d say I touch a Windows box for about 5 minutes a month, and usually just to use a browser. A few weeks ago, I was asked to install Windows on a desktop for someone. I did. I then attempted to install Firefox. Using what I remembered of Windows, I navigated to the “Control Panel” and clicked (err… double clicked) on “Add and Remove Programs”. Seems logical enough. I then stared at the screen for about 30 seconds, trying to find the Search box, where I could type in “Firefox”. Finally, I literally began laughing out loud, when I remembered that Windows doesn’t have unified package management, and I’d need to manually find the Firefox binary on their web site, download it, and run whatever installer program Firefox chooses to use. Same issue with updating software. I’m utterly perplexed, being a Linux user, that Windows and Mac people still search through Google or multiple web sites just to find new software. I’m even more perplexed that the OS update/patch program doesn’t also update all of the software on the system. It seems like the stone ages.

In my opinion, one of the biggest failings of modern Linux package management is the assumption (derived from multi-user systems) that all software should be installed system-wide. Granted, it doesn’t do a whole lot to actually protect a single user if they install malicious software available to just themselves (especially since most desktop installs these days are probably used as single-user systems), but I really feel that distros (especially desktop-oriented distros) should have an option to easily install packages for just the current user, and possibly do this by default.

I can’t find the link right now, but I did find an interesting article on Microsoft’s old anti-Linux campaign (“get the facts”). One of the things mentioned was that when Microsoft compared “vulnerability counts”, they were actually comparing: 1) entire Linux distros vs just the core Windows OS, and 2) counting individual patches in Linux versus patch sets released by MS. So, not only was MS literally counting apples and oranges, but they were totally ignoring unfixed vulnerabilities. Given Microsoft’s habit of not fixing vulnerabilities - especially in “unsupported” products - it’s no wonder how they got the numbers to look so good.

So, here’s a thought. People are used to paying for an OS and for software. Start a Linux vendor that sells a desktop, newbie-oriented Linux distro. Charge a per-user flat rate for the distro and a bunch of base packages, that includes X hours of telephone support. Charge per hour/minute/whatever for additional support. Bundle in secure VNC, secure remote access, etc. in a way that will allow support to remotely access the computer, but preserve the privacy and security of the user (perhaps an app that allows the user to initiate a reverse VNC or SSH session to support). Lock down root access - allow the user to do it, but remind them every time that, outside of a specified set of commands, their actions will be logged and won’t get full support. Then figure out a way for support to write a shell script that’s sent to the user to perform administrative actions, which will all be listed in relatively simple terms for the user to examine and approve. Finally, have a giant package repo, all of which is free or comes with paid support. Any F/OSS packages that aren’t already in the repo can be requested by a customer, and for a flat fee for the first requesting customer (say, $10) will be examined, approved, packaged, and added to the repo.


comments powered by Disqus