Slashdot Mirror
Ask Slashdot: Security Digests For the Home Network Admin?
New submitter halcyon1234 writes "I'm currently cutting the webhost cord, and setting up a simple webserver at home to host a couple hobby websites and a blog. The usual LAMP stuff. I have just enough knowledge to be dangerous; I know how to get everything set up and get it up to date, but not enough to be sure I'm not overlooking common, simple security configurations. And then there's the issue of new vulnerabilities being found that I'm not even aware of. The last thing I want is to contribute to someone's botnet or spam relay. What readings/subscriptions would you recommend for security discussions/heads up? Obviously I already read (too much) Slashdot daily, which I credit for hearing about some major security issues. Are there any RSS feeds or mailing lists you rely on for keeping up to date on security issues?"
When you're done with your setup. Post a story on Slashdot linking to your website, that's a fairly good stress test.
/.
Bonus points if you add something like "My awesomely new bulletproof website!". That should kick off the reliability test engines from
http://www.securityfocus.com/
Most ISPs do NOT allow this kind of stuff. While it might fly under the radar, there is always the possibility they will shut off your access. Besides, with a dynamic IP any change to it will take your website offline until DNS catches up. Hosting is cheap, I don't see why you'd want to cancel it unless it's hurting the bank.
You may (or may not) know to do these things:
Only allow SSH pubkey access for maintenance. Turn off remote passwords altogether in sshd_config.
Move SSH to some non-standard port (away from 22). Just makes it harder for outsiders to find the port.
Run something like denyhosts to kick off people trying brute force passwords on your server and to log their activity.
Consider only allowing SSL (port 443) access to the webserver and close down all other ports in iptables.
Just a few things (which may be obvious) to get you started.
You said LAMP--well, most L distros have a security list you can subscribe to to keep up-to-date on this sort of thing. Also, Linux Weekly News (lwn.net) regularly posts security announcements from most major distros
I am running a LAMP server (only the LA part active) with a few Web pages on my obsolete home desktop, with a slow ADSL connection. I don't have anything of serious value on the machine, so presumably an attacker would only benefit by using me as a bot. The system logs show regular access by the indexing services: Google, Yahoo, ... and lots of what looks like simple-minded dictionary attacks, and some probes for SQL/PHP features that are not available. None of the login attacks appears to have found my user name, much less hit the password. I use a long password, generated by a random generator. That seems to be enough.
There is some chance that I am owned by someone I can't detect. But I never notice much activity on the CPU nor on the network, and my ISP doesn't complain that any bad behavior comes from my connection. If I suspected a problem, I would power off while considering my response.
I have tripwire installed, but every time I try to check it I get confused by the pass-phrase management and put it off.
Mike O'Donnell http://people.cs.uchicago.edu/~odonnell/
Where's the MyCleanPC guy when you need him?
Forget running it at home, get an Amazon AWS Micro site; it's free and the default config is as secure as possible. IPTABLES is easy to configure via the AWS web gui or you can manually create better rules once on the system. AMZN keeps track of updates, you just have to remember to ssh in every once in a while.
some sites:
http://www.securitywizardry.com/radar.htm
(a little heavy on the java)
https://isc.sans.edu/
You could subscribe to the CERT messages, but they kinda lag. There are some good security related mail lists which
I can't remember at the moment...
Check available updates for packages and kernel...
Look at mod_security for apache
If you're running wordpress or some other CRM app, be careful on how much you rely on third party packages
If you have phpadmin or webadmin installed, you may want to limit what IP's have access to it.
If you're running sshd, you may want to block bruteforce attempts after a certain number of bad tries, You should
probably just use certificate based authentication instead of passwords.
The best place to start is here
http://www.us-cert.gov/cas/signup.html
then onto the security announce list of whatever distro you use.
Those two alone will probably give you enough information to keep your system safe
Normal people worry me!
First: The only way to connect to your system is over a logical port. So, learn netfilter / IPtables and shut down all ports you don't need. The book "Running Linux" by Dalheimer and Welsh has a pretty good section on netfilter / IPtables. My recommendation - just leave port 22 and 80 (maybe 443 if you're having people log into your web application remotely). Default policy is drop packets unless it matches one of those ports.
Second: Turn off remote root login, typically found in sshd_config. This'll stop much of the probing.
Third: You don't want to allow someone to relentlessly try passwords. Get a program like Fail2ban. This will allow a certain number of login attempts before it bans the IP, just dropping the packets and not letting the password authentication module test them.
Fourth: Strong username/password combinations. The attacker has to guess the correct combination. Get jiggy with it. Unusual username and unusual passphrase password. Especially for the root user.
Fifth: Stop having Apache broadcast all of its version information. When someone is looking at response headers, they should see just that it's Apache and not Apache version XYZ. Apache loads several config files and reads them as one long config file (they're broken up for easier management). There's a setting in Apache to do that.
Sixth: In Apache's config files, turn off directory listings. Again, a simple configuration text file setting which eludes me at the moment. Apache The Definitive Guide by Laurie and Laurie is a good book to have. This info is also available on the web.
Seventh: Read your log files regularly. auth.log, error.log are very informative ones. Doing a lastlog command on a regular basis helps.
Finally - What is security?
1) You don't want people writing to where they shouldn't be writing.
2) You don't want people reading what they shouldn't be reading.
3) You don't want people executing what they shouldn't be executing.
Set up permissions well. Don't change them willy-nilly but if reading/writing most stuff on your box requires being part of the root group, that's pretty good security.
Finally, finally - keep reading various technical sites on the web for new security problems. Address as necessary.
On a publicly visible web server is to set up set the directive for the default web site (the first one in the virtual host list) to default deny to everyone. Then put your web site on a different virtual host. 99.9% of the scans I see come in by IP address, which gets them the default site. Any legitimate traffice will come in by domain name. This set up not only denies the script kiddes access to any PHP forms you've got, it convinces their 'bots to give up very quickly, which means less of a toll on your bandwidth.
(As someone noted, the standard consumer highspeed account prohibits running servers. Many commercial accounts do, too, unless you told them you're running a server of some kind. You may also have to get them to unblock port 25 if you want to run your own mail server - be very careful if you do that, though. You don't want to be a spamfest rathole without knowing it.)
Your distro will have a regular patch channel that will address most vendor-introduced vulnerabilities. Patch religiously, and often. At least once per week. It's not like you're responsible for SLA's or regression testing. If you somehow uncover a bug when you patch, muscle through it, and keep going.
Use a firewall and only expose necessary ports. Protect the ports with strong authentication, encryption where applicable, and possibly a reactive blocker such as fail2ban to keep the script kiddies at bay. If you must run an external SSH server, run it as a seperate process, and only allow key auth, and only for a single user.
Get on whatever mailing lists or errata lists support your distro and apps, and try and keep up with them. If your apps are maintained as source, try and use the repos to update your apps instead of just relying on standard stable packages. You'll get bug fixes faster (probably bugs as well. See above)
Use something like logwatch and read the daily mails.
Also use something like rkhunter to alert you in case something changes.
-- lk t lv ll th vwls t f wrds. T svs lts f tm t wrt bt ts pn n th ss t rd nd mks m lk lk cmplt dpsht.
It's called "staging".
F*ck comments. F*ck all the other interactive "web 2.0" sh*t. Do your Wordpress or whatever, then suck it out of the DB, convert to static HTML, and put it on the external webserver.
Problem (pretty much, well 99%) solved.
My company supports about 700 Debian servers that are running at customer sites; as far as I can tell, not a single one of these has been compromised due to an unpatched security vulnerability in one of the Debian packages. Configure some kind of auto-update (e.g. cron-apt); Debian is by default configured to only install security updates and these are usually released within one or two days.
We did have a few servers that were compromised due to customer mistakes (unsafe root passwords especially) though. From what I've learned in these incidents, you can spot an infected machine by these symptoms:
In theory, a rootkit could be hidden really well, but in practice, the attackers just don't seem to care to adapt their rootkits specifically to the distros. An experienced Linux admin should have no problem detecting them.
Most ISPs offer business accounts that you can do whatever you want(to a degree).
Bugtraq and Full Disclosure mailing lists are a good read. Almost all new vulnerabilities are posted to one of these lists. In addition, many Linux distros post their security notices here (Ubuntu used to, but now only posts on their own list). The CERT list mentioned by previous commenters is also good, even if it can be a little slow at getting the news out. Microsoft, Apple and others report their security notices through this list.
"Every man has a right to his own opinion, but no man has a right to be wrong in his facts." - Bernard Baruch
Problem is, once it gets to your router, it's too late, your bandwidth is already gone.
I've been using a VPS for $3/month from 123systems.net. I haven't done much with it yet, and I don't know how consistent it is, but so far I have no complaints. buyvm.net was another I was looking at that I believe has an even cheaper option ($15/yr!). Like someone else said, check out http://www.lowendbox.com/ to become informed about the options. Of course, you get only a pittance of ram/cpu for these bargain basement prices (and often limited availability -- buyvm sounds like a bit of a lottery), but it is still nice to have full control over a linux system that I can pack it up and deploy it to another linux server with more resources/consistency if/when I need to, while playing around with it for cheap now. It's also nice to have a far away offsite backup in case my city gets EMP'ed / destroyed by aliens / etc.
Also, like someone else mentioned, I have run ssh/www for about 15 years on my home ISP since whenever I got broadband with no complaints from my ISP.
That depends, your server can just as easily be set up not to respond. Basically if you are trying to save resources that attackers are using, you either need your upstream provider to do it, or it really won't matter if your router or your server do the blocking. The only exception would be if your internal link from your router to your server is congested, but if that's the case than you really have bigger issues.