Payload: the virus could do something as simple as expose a phpinfo()
on the host, which could then be used to assess what else could be done on the host, based on it’s setup. Or it could be a DDoS node by using the server to infinitely request a target server. Or it could wipe files on the host system. Or, indeed, it could be setup in a way that it could take an arbitrary payload that could be executed at will on the host.
Self-replicating: you could have a script that inserts itself in the head of other PHP files on execution, and then those files are able to replicate their malicious payload into other files themselves.
Avoid detection: if you are going to be doing something that calls eval()
or similar, then you’ll want to do something to make sure that string doesn’t appear in your code. You could do a number of things to achieve this, which will not be discussed.
Take advantage of weaknesses: Along with eval()
, include()
is in this category, as is base64_decode()
. You really shouldn’t ever be using these things, and you should be wary of anything that does. If you think that using one of these functions is a good idea, you may need to reassess what you are trying to achieve.
“eval() is nasty. Unless you are writing viruses. Then it’s pretty cool.” - @bendechrai.
Resource stealing via host analysis: taking advantage of a machine enables someone that isn’t authorised to use its resources for nefarious purposes. If the machine is a beefy server, it could have a large amount of disk that could be used for seeding torrents, excess memory could be re-purposed for crypto-currency mining. A large network could be used for the spreading of itself, or as a node point for interrogating other infected machines and siphon data.
Obfuscation and cryptography: the code could implant itself in other files by signing itself with a one time key, and then unlocking itself using that same key. Using this method, any future copies of itself will not match - the encrypted version, each using a different key, will produce a different output - so therefore wouldn’t be very noticeable.
Lay dormant, and move slow: a process editing a hundred thousand files in a second on the one host is detectable by modern intrusion detection systems, and vigilant system administrators. A process that edits one file, waits a bit, then edits another, is less noticeable.
Long line possible attack vector: you could, as a developer attempting to mitigate against all the possible vectors listed above, use a lot of different file upload sanitation techniques, such as checking host headers, ensuring a valid file name, ensuring that the uploaded file is loaded from the correct directory, and that the mime type and content-type are good. But if you just include()
the image, then you’re going to have a bad time. In the demonstration, obfuscated malicious PHP code was inserted into the sample JPEG, just under the header that the mime-tye checkers use to validate the file type. By running include, the embedded PHP code is executed. Game over.
Mitigation: Don’t use eval()
. Don’t use include()
. Don’t use base64_encode()
. Be very wary of any code you see that employs these. Also, any function blocks that appears to be very badly coded that might use string transformations to build strings like this. Keep an eye out for anything without spaces, or that you cannot understand by sight. Sanitise any input you get. Never trust input, even if it’s from other servers within your network. Never trust user input, and be very cautious of any uploaded files.
Be wary that attack vectors are found, mitigated against, and evolve. The things you are protecting against today won’t be what gets you popped tomorrow.
I was fortunate enough to be able to see this presentation live.
I know Ben has a very engaging speaking style, but I have never seen a presentation like this before. There were some technical issues that made the talk stall a bit: Ben’s own laptop wasn’t able to be used with the conference hall projector; he attempted to use my own laptop, but the HDMI cabling at the front of the room wasn’t working, however, the one in the back of the room was. He was able to use a third laptop, but because of the live demonstration, not all the prerequisite components were installed, and some live bug fixes had to be completed during the presentation.
However, this was at PHP Australia. The room was full of people who develop PHP, and even the original creator of PHP and some of the current core developers. Throughout the talk people were suggesting bug fixes, apache config changes, packages to install and tests to try and get things working. I have never been in such an environment, and I doubt I ever will.
The entire talk was an experience, even with the little niggles, which Ben took in his stride to present something that I will never forget. It was one of the highlights of the entire conference.