There will never be a turnkey system for real security, there will just be a "club" like device. The automotive anti-theft device "the club" works on the premise that its easier to steal someone elses car than yours. It is not hard to remove (most steering wheels are plastic and easily cut) but it takes time and effort. Sure if someone really wants into your device they can get in but if you make the hurdle high enough they will just move on to someone elses device. Off the shelf malware/av scanners raise that bar to stop the casual attacker, they do not stop the determined attacker.
Security and usability is always a tradeoff. You cannot have a lot of one without reducing the other. You could for example set the write protect switch on the SD card and work solely off tempfs partitions but that greatly decreases usability. Periodic reboots would mean that anything installed is purged.
You must also define "secure from whom." A vending application where the device exists somewhere outside a secure facility (such as your home) can be compromised differently than one that is locked up somewhere. Physical security is just as important for certain types of threats. The steps you take should be largely determined by the realistic threat model for your specific environment and data. I suggest a paper on "Attack trees" which you can read here
https://www.schneier.com/academic/archi ... trees.html
If you are worried about random internet things, like the millions of scanners out there right not actively looking for devices with a username of "pi" and password of "raspberry" then you can do a couple things. 1. disable ssh (which is the default since last Nov). 2. if enabled make sure the password is changed. 3. try putting ssh on an alternate port (most of the scanners are really dumb just connecting to port 22 and trying a bunch of user/pass combos until one works or the list is exhausted). 4. look for "honey pot" software that will trigger alerts if accessed. Could be a simple daemon on a port and if your box is portscanned it will alert you that someone is looking (be ready for a slew of alerts that are better ignored). Honey pots can also be quite complex depending on need (Pi users probably do not have that high of a need on their devices).
If you are worried about browser related attack vectors there are plugins for browsers that are platform independent. You must of course run a compatible browser.
For malware/av you can look for vendors, probably specializing in mobile devices, that do various things. Although many will require things like an android base layer and wont scan much outside of that environment so look carefully. Stuff does exist generically for linux like
https://www.rfxn.com/projects/linux-malware-detect/ which can be compiled if arm binaries are not available and then periodically run. Of course if there is a rootkit successfully installed that rootkit can mask itself and never be seen. To scan for such things you would need to get a clean known install and build a DB of what the files look like and then pull the card, mount it RO somewhere and scan it. Tripwire is one such application that is designed to accomplish this task.
Threats evolve, what is secure today may not be secure tomorrow. You can also periodically scan your Pis with tools, several are available.
http://www.openvas.org/ is open source as is metasploit (but I think openvas is easier for the non-tech people to use).
Then there are a slew of network based scanners and filters. These would be installed between your Pi and the network, they will monitor network traffic for suspicious activity. Snort is one such tool but there are many. Proxies can also provide web filtering. However they may miss things like embedded stuff in PDF documents, word documents (OOO does support some of this), etc.
You can also look into tools like systrace which you can read more about here
http://niels.xtdnet.nl/papers/systrace.pdf Basically its a wrapper that will prevent system calls (file open, socket open, etc) that are out of scope for the app. So if someone does manage to exploit an application, lets say a web server, they are unable to access files outside of /var/www or whatever. They are unable to open new ports say for a reverse shell. Its not 100% but it is one of many components that can help to harden a system.
I guess the question is how far do you want to take this?
