Difference between revisions of "Forensic"
Line 33: | Line 33: | ||
Collect data about the system's state (consult the manpages if you are unsure about what you are doing): | Collect data about the system's state (consult the manpages if you are unsure about what you are doing): | ||
<pre> | <pre> | ||
#------------- | #------------- | ||
mkdir incident_data | mkdir incident_data | ||
Line 51: | Line 50: | ||
for i in t p c t l; do ipcs -a -${i} > ipcs_a_${i}.txt;done | for i in t p c t l; do ipcs -a -${i} > ipcs_a_${i}.txt;done | ||
#------------- | #------------- | ||
</pre> | </pre> | ||
If there are suspicious processes that need further analysis, preserver the original binary and dump the program's memory: | If there are suspicious processes that need further analysis, preserver the original binary and dump the program's memory: |
Revision as of 23:56, 23 May 2011
| Mission | Members | Contacts
| Incident handling | Alerts | Monitoring | Security challenges | Procedures | Dissemination
FORENSIC HOWTO
You may want to see egi incident response procedure at : [ https://wiki.egi.eu/wiki/EGI_CSIRT:Policies]
Release of 19 may 2011, edited by Heiko Reese <Heiko.Reese(at)kit.edu>
Linux Forensics HowTo
This document is a constant work-in-progress. Comments and additions are always welcome.
This document describes a best-effort approach for preserving and analyzing compromized Linux installations. Because there are many different Linux userlands (aka distributions), some commands may require a different syntax or different commands (most notably when package management is involved) to achive the same goal. To follow the instructions in this document, at least a basic understanding of the procedures presented here is necessary.
Forensic analysis consists of (at least) these phases:
* Identify the system. * Gather data. * Analyzie the data.
Identify compromized systems
TODO: (logs, monitoring, netflows, suspicious or erratic behaviour, external notification, etc...)
Gather data
The data aquisition process is twofold: first, gather information from the running (live) system. After that, analyze the »cold« system.
If the system runs as a virtual machine, freeze/pause it and create dumps/images from the filesysems/blockdevices and the memory.
Try not to write to the local filesystem. Put all gathered data onto external drives, network shares or into a ramdisk.
Collect data about the system's state (consult the manpages if you are unsure about what you are doing):
#------------- mkdir incident_data cd incident_data ps -auxwwwe > ps_auxwwwe.txt netstat --program --notrim --verbose -n > netstat_pTvn.txt netstat --program --notrim --verbose > netstat_pTv.txt w > w.txt last > last.txt lastlog > lastlog.txt cat /proc/mounts > proc_mounts.txt arp -n > arp_n.txt ip neigh show > ip_neigh_show.txt ip route list > ip_route_list.txt ip link show > ip_link_show.txt lsof -b -l -P -X -n -o -R -U > lsof_blPXnoRU.txt for i in t p c t l; do ipcs -a -${i} > ipcs_a_${i}.txt;done #-------------
If there are suspicious processes that need further analysis, preserver the original binary and dump the program's memory:
{{{ #------------- export PID=12345 # <- INSERT PROCESS-ID (PID) HERE kill -STOP ${PID} # stop process cp /proc/${PID}/exe ${PID}.exe # some distributions have a script called "gcore" which does this in batch-mode gdb -p ${PID} # type "gcore", then "detach" and "quit" # The program's memory is now saved as core.PID. ls -l /dev/shm # look for shared-memory-segments owned by the process # by doing grep '/dev/shm' /proc/${PID}/maps # copy them if deemed neccessary tar cvf proc_${PID}.tar /proc/${PID}/{auxv,cgroup,cmdline,comm,environ,limits,maps,sched,schedstat,sessionid,smaps,stack,stat,statm,status,syscall,wchan} kill -9 ${PID} # kill process #------------- }}}
Create a list of all files in the system:
{{{ #------------- mkdir /mnt/root_ro mount --bind / /mnt/root_ro mount -o remount,ro /mnt/root_ro # do not combine the two previous steps, this won't work on some older kernels find /mnt/root_ro -xdev > find_root_ro_xdev.txt umount /mnt/root_ro #------------- }}}
Install/copy chkrootkit (http://www.chkrootkit.org), rkhunter (http://rkhunter.sourceforge.net) and ossec-rootcheck (http://www.ossec.net/main/rootcheck) to the machine.
Remount all „real“ filesystems as read-only ({{{mount -o remount,ro MOUNTPOINT}}}). This is best done manually by the administratior. You may use this very simple heuristic as an alternative if needed:
{{{ #------------- sync for mountpoint in $(sort -r /proc/mounts | grep -E ' (ext[234]|xfs|reiser|vfat|ntfs)' | cut -d' ' -f2) do echo mount -o remount,ro "${mountpoint}" done #------------- }}}
Run chkrootkit, rkhunter and ossec-rootcheck.
Some package management-systems have checksums of their installed packages. Debian-based systems offer {{{debsums}}} and Redhat has {{{rpm -Va}}}. Save the output.
Copy the collected data someplace save or remove/unount the external storage/network drive.
Do not shutdown as usual! Disconnect the power the the system.
Remove the harddisks and create images (use http://www.gnu.org/software/ddrescue/ddrescue.html).
Data analysis
In order to proceed from this point on, answer these questions:
* What do you already know? How can you use this knowledge to proceed? * What is your next goal? Finding the breakin point? Understand malicious code/backdoors? Find culprits? Identify other systems involved? * Do you have the resources/manpower to analyze now? Or should these resources be used to mitigate the threat (assuming that is still exists)?
TODO: (explain procedures for standard problems)
Here are a few pointers to helpful software and services:
* TODO: malware-foo * TODO: network intelligence * TODO: binary analysis