Methods of Shared Web Application Server Compromise
Pretty much endless. Have a look at the Open Web Application Security Project's 2013 Top 10, (www.owasp.org) just for starters. If you're anything like me, you're already trembling.
Threat and Risk Assessment: Scope
A full threat and risk assessment would also take into account compromise of user authentication data, e.g. cross-site request forgery, because one compromised session could permit unrestricted access to the server if the account is not isolated properly. For example, consider a hacker exploiting a known vulnerability in a web application after having successfully hijacked a valid user's session.
What you're asking about is something that is more often considered in the highest-security scenarios: banking, finance, government, military... And in all of those cases servers with high-value assets are almost categorically isolated and not shared. The evaluation of what is high value is often made on the basis of three principles: confidentiality, availability and integrity.
Example of classic industry high-security scenarios
For example, for a bank, all the financial transaction data is highly confidential, very restricted in availability and its integrity is paramount. That's why most banks will isolate their server clusters by application and by relative interactivity. Then they will also make routine redundant backups of the information into geographically disparate locations, on equally secure clusters.
The best methods to prevent server compromise are routine hardening:
- the strongest passwords combined with multifactor authentication,
- secure-signed kernels and software only, e.g. SELinux
- strict controls using a software firewall like
iptables defining which applications are allowed to use which ports in what direction and by what users, but also hardware-based firewalls,
- active monitoring of the network traffic using a software like Wireshark6,
- you can also use [intrusion detection and prevention systems, possibly even two or three such systems at the same time,
- if you are safeguarding high-value information that is exposed to high risk (according to your threat and risk assessment), then physical security should also be considered as a fundamental factor (i.e. who has access to the datacentre, how are they cleared for access, how are keys managed),
- you can find more information from NIST Special Publication 800-53 -- Anecdotally, I remember hearing a statistic that NASA had only implemented about 65% of these suggested controls by 2014, by virtue of how exhaustive they are.