mirror of
https://github.com/jackyzha0/quartz.git
synced 2025-12-27 14:54:05 -06:00
vault backup: 2022-09-14 21:19:35
This commit is contained in:
parent
4d11e44c1c
commit
15bd07bd64
@ -70,4 +70,54 @@ Once score is calculated, the course of action can be determined using the follo
|
||||

|
||||
|
||||
### Security through obscurity
|
||||
an optimistic/idealist approach. Better to think of STO as a
|
||||
an optimistic/idealist approach. Better to think of STO as a pseudosecurity measure
|
||||
|
||||
### Shannon's Maxim
|
||||
"The enemy knows the system", i.e., assume that any security mechanism is not secret.
|
||||
|
||||
### Kerckoff's Principle (crypto specific)
|
||||
A cryptosystem should be secure even it its inner workings (though not the key) are know to an attacker
|
||||
|
||||
### Questions to consider
|
||||
- what are the stakes?
|
||||
- why might you be a target?
|
||||
- who are your potential adversaries?
|
||||
- what resources might they have?
|
||||
- what is the cost to safeguarding against plausible attacks?
|
||||
- can you put a monetary value on your data or uptime?
|
||||
|
||||
# Intrusion detection
|
||||
Knowing that an attack has happened, who did it etc
|
||||
|
||||
- "logging in"/"logout" the event is logged
|
||||
- log files get large to rotation (splitting, compression and culling) is used
|
||||
- need to have a policy of retention and level of detail for log files
|
||||
- unix-like systems tend to use a standard format (plain test, one line per event, with a timestamp)
|
||||
- may need specialised tools to searching logs for relevant events
|
||||
- correlations, etc
|
||||
|
||||
many levels
|
||||
- system wide events (startup/shutdown)
|
||||
- often includes memory dump
|
||||
- authentication events (log in/log out, sudo commands)
|
||||
- network level (esp in conjunction with firewall rules)
|
||||
- log attempts to access ports
|
||||
- service specific (e.g., web server, db server, SSH)
|
||||
|
||||

|
||||
client ip, timestamp, request, response code, data length
|
||||
|
||||
## system monitoring in general
|
||||
attacks will often incur ususual loads on the system
|
||||
- more CPU usage
|
||||
- high login attempt frequency
|
||||
- large number of running processes
|
||||
- (unusuallly) high memory use
|
||||
- unusual or high network activity (DDoS or incoming attack)
|
||||
- unusual disk/io usage
|
||||
|
||||
standard tools
|
||||
- top
|
||||
- busiest processes - and their information (CPU, Memory)
|
||||
- general system information (cpu usage, num processes, network, etc)
|
||||
-
|
||||
Loading…
Reference in New Issue
Block a user