Lecture 4: Unix Security Basics

Asoc. Prof. Guntis Barzdins

Asist. Girts Folkmanis
University of Latvia
Oct 8, 2004
 Top UNIX Vulnerabilities

 U1 BIND Domain Name System  U6 Sendmail

 U2 Remote Procedure Calls (RPC)  U7 Simple Network Management
 U3 Apache Web Server Protocol (SNMP)
 U4 General UNIX Authentication  U8 Secure Shell (SSH)
Accounts with No Passwords or  U9 Misconfiguration of Enterprise
Weak Passwords Services NIS/NFS
 U5 Clear Text Services  U10 Open Secure Sockets Layer
(SSL)

Source: http://www.sans.org/top20/#threats
 Favourite TCP Ports
 20 FTP (data)  7-19 echo, discard, daytime, chargen, netstat
 22 SSH
 21 FTP (control)
 42 wins
 23 Telnet
 53 dns
 25 SMTP (mail)  111 sun rpc
 70 Gopher  113 identd
 79 Finger  123 ntp
 80 HTTP also 8000 or 8001 or 8080  135 loc-srv/epmap – used to attack wintel
 110 Pop3  137-139 netbios
 119 NNTP (news)  161 snmp
 143 Imap  512-517 rexec, rlogin, rsh, talk, syslog, who
 635 mountd – Linux
 2049 nfs
 6670 Deepthroat
 31337 BackOrifice
 No system is perfectly secure,
but still we need security
 A number of toolkits exist that allow total amateurs
to become holy terrors.
 The good news is that if you can beat the popular
intrusion toolkits, 90 percent of the bad guys will
go bother somebody else who's less secure.
 Protection

 Operating system consists of a collection of objects,

hardware or software

 Each object has a unique name and can be accessed

through a well-defined set of operations.

 Protection problem - ensure that each object is accessed

correctly and only by those processes that are allowed to
do so.
 UNIX Security Basics

 Permissions
 UID
 GID
 Superuser
 SUID, SGID
 Sticky bit
 Umask
 Filesystem restrictions
Domain Implementation in UNIX

 Two domain groups

 User
 Superuser (can do everything, UID=0)
 User domain group
 Domain = user-id (UID)
 Domain switch accomplished via file system.
 Each file has associated with it a domain bit (setuid bit).
 When file is executed and setuid = on, then effective user-id
is set to owner of the file being executed. When execution
completes user-id is reset (exit() for child process ).
 Subjects and Objects

Each subject (process) and object (file, socket, etc)

has a 16-bit UID.
 Each object also has a 16-bit GID and each subject has one or
more GIDs.

Objects have access control lists that specify read, write,

and execute permissions for user, group, and world.

Super-users (uid=0 root) can do anything.

 Subjects and Objects
Objects = files (regular and devices /dev)
UID GID Others
(effective UID, GID counts)

UID User
permissions
Subjects = processes

GID-main+ Group
permissions
GID-list
Others Others
permissions
 inodes
 inodes contain a lot of information about a file
 mode and type of file
 number of links to the file
 owner's UID
 owners GID
 number of bytes in file
 times (last accessed, modified, inode changed)
 physical disk addresses (direct and indirect blocks)
 number of blocks
 access information
Unix File System (UFS) Structure
 Directory
 Under UNIX directories are special (OS writable only) files.
 The directory file is an unsorted linked list of filenames to file-inode
(attributes and location of file on hard disk)
 Directory size will always increase to be large enough to hold all
the file entries. If the number of files latter shrinks the directory
size WILL NOT!

5 apples
4 oranges
5 aboli
2 .
7 ..
 ls -l
> ls -l foo
-rw-rw---- 1 hollingd grads 13 Jan 10 23:05 foo

permissions size name

owner group

time
 File Time Attributes
 Time Attributes:
 when the file was last changed ls -l
 when the file was created* ls -lc
 when the file was last read (accessed) ls -ul

*actually it’s the time the file status in the directory

last changed (e.g. file renamed).
 Types of Files

 Regular Files
 binary
 GIF, JPEG, Executable etc.
 text
 scripts, program source code, documentation
 Supports sequential and random access
 Types of Files (cont.)

 Directory
 Can contain ANY kind of files
. (Dot) The special name for the current directory.
.. (Dot) (Dot) The special name for the directory above
the current directory.

 Device File
 Allows programs to communicate with hardware.
 Kernel modules handles device management.
 Types of Files (cont.)

 Device Files (cont.)

 Character Device
 Accepts a stream of characters, without regard to any
block structure.
 It is not addressable, therefore no seek operation

 Block Device
 Information stored in fixed-sized block
 It is addressable, therefore seek operation is possible.
 Types of Files (cont.)

 UNIX Domain Sockets (BSD)

 sockets that are local to a particular host and are
referenced through a file system object rather than a
network port.
 X windows
 Named Pipe
 Allow processes to communicate with each other.
 Types of Files (cont.)

 Hard links
 Linking files by reference
 System maintains a count of the number of links
 Does not work across file systems.
 Soft links
 Linking files by name
 No counter is maintained
 Work across file system
 From “man ln”
 There are two concepts of `link' in Unix, usually called
hard link and soft link
 A hard link is just a name for a file. (And a file can have
several names. It is deleted from disk only when the last name
is removed. The number of names is given by ls(1). There isno
such thing as an `original' name: all names have the same
status.
 A soft link (or symbolic link, or symlink) is an entirely different
animal: it is a small special file that contains a pathname.
 Creating a Link

 Create a link directory by typing the following

command from your home directory:
soft link
% ln -s /home/faculty/ostic/prof myprof

 You only need to create this link once. It will

appear as a subdirectory in your home directory
structure every time you log on to the system.
 Disk vs. Filesystem
 The entire hierarchy can actually include many
disk drives.
 some directories can be on other computers

bin etc users tmp usr

hollid2 scully
 Disk mount options

 Override individual file permissions

 A major security tool in Unix
 File permissions
File type
- : plain file
d : directory
c : character device (tty, printer)
b : block device (disk, CD-ROM) Access granted to
l : symbolic link
s : socket others
=, p : FIFO

-rwxr--r--
Access granted to owner Access granted to
r : read / w : write / x : execute group member
 Permissions for Files

 If you have read permission for a file, you

can view its contents.
 If you have write permission for a file, you
can alter its contents.
 If you have execute permission for a file, you
can run the file as a program.
 Permissions for
Directories
 If you have read permission for a directory,
you can list the contents of the directory.
 If you have write permission for a directory,
you can create or remove files or directories
inside that directory.
 If you have execute permission for a
directory, you can change to this directory
using the cd command, or use it as part of a
pathname.
 SUID/SGID/sticky bits
 SUID (set uid)
 Processes are granted access to system resources based on
user who owns the file.
 SGID (set gid)
 (For file) Same with SUID except group is affected.
 (For directory) Files created in that directory will have their
group set to the directory's group.
 sticky bit
 If set on a directory, then a user may only delete files that he
owns or for which he has explicit write permission granted,
even when he has write access to the directory. (e.g. /tmp )
 File Permissions
 File Permissions (ex: rw-r--r--)
 owner: rw-, group: r--, others: r--
 r: read, w: write, x: execute
 When a process executes, it has four values
related to file permission
 a real user ID, an effective user ID
 a real group ID, an effective group ID
 When you login, your login shell process’ values are
your user ID and group ID
 Effective User and Group ID
 A process’ effective user ID
 depends on who executes the process, not who
owns the executable
 E.g., if you run passwd (owned by root), the effective
user ID is your ID, not root; then how can it update
/etc/passwd file owned by root ?
 Two special file permissions
 set user ID and set group ID
 When an executable with set user ID permission is
executed, the process’ effective user ID becomes that of
executable; the real user ID is unaffected
 File permission of /bin/passwd is r-sr-sr-x
 Real uids
 The uid of the user who started the program is used as its real
uid.
 The real uid affects what the program can do (e.g. create,
delete files).

 For example, the uid of /usr/bin/vi is root:

 $ ls -alt /usr/bin/vi
lrwxrwxrwx 1 root root 20 Apr 13...

 But when I use vi, its real uid is dkl (not root), so I can only
edit my files.
 Effective uids
 Programs can change to use the effective uid
 the uid of the program owner
 e.g. the passwd program changes to use its effective uid
(root) so that it can edit the /etc/passwd file

 SUID bit enables this functionality

 Real and Effective Group-ids

 There are also real and effective group-ids.

 Usually a program uses the real group-id

(i.e. the group-id of the user).

 Sometimes useful to use effective group-id

(i.e. group-id of program owner):
 e.g. software shared across teams
 SGID bit enable this functionality
 Sample SETUID Scenario
 /dev/lp is owned by root with protection rw-------
 This is used to access the printer
 /bin/lp is owned by root with rwsr-xr-x (with SETUID=1)
 User A issues a print command
 Shell (running with A’s UID and GID) interprets the command and forks off
a child process, say, P
 Process P has the same UID/GID as user A
 Child process P executes exec(“/bin/lp”,…)
 Now P’s domain changes to root’s UID
 Consequently, /dev/lp can be accessed to print
 When /bin/lp terminates so does P
 Parent shell never got the access to /dev/lp
 File system tips

 Turning off SUID / SGID in mounted file system

 use nosuid (and nodev if possible) when mounting
remote file system or allowing users to mount floppies
or CD-ROMs
 Finding SUID and SGID Files
 # find / \( -local -o -prune \) \( -perm -004000 -o -perm
-002000 \) -type f -print
 ( xdev can be used in place of local/prune)
Unix Accounts and the Filesystem
 Unix Accounts

 To access a Unix system you need to have an

account.
 Unix account includes:
 username and password
 userid and groupid
 home directory
 shell
 Creating user accounts

 useradd or adduser scripts

 manually
 edit /etc/passwd, etc/shadow, etc/group
 remember to lock these files while editing - vipw
 run “passwd [user]”
 create home directory
 chown, chgrp, chmod
 copy defaults (e.g umod) from
 /etc/skel
 /etc/profile
 username
 A username is (typically) a sequence of alphanumeric
characters of length no more than 8.
 username the primary identifying attribute of your
account.
 username is (usually) used as a part of email address
 the name of your home directory is usually related to
your username.
 password
 a password is a secret string that only the user
knows (not even the system knows!)
 When you enter your password the system
calculates a hash (one-way) function and
compares it to a stored string.
 passwords are (usually) no less than 8
characters long.
 It's a good idea to include numbers and/or
special characters (don't use an english word!)
 userid

 a userid is a number (a 16-bit integer) that

identifies a Unix account. Each userid is unique.
 It's easier (and more efficient) for the system to
use a number than a string like the username.
 You don't need to know your userid!
 Unix Groups and groupid
 Unix includes the notion of a "group" of users.
 A Unix group can share files and active processes.
 Each account is assigned a "primary" group.
 The groupid is a number that corresponds to this
primary group.
 A single account can belong to many groups (but
has only one primary group).
 Home Directory

 A home directory is a place in the file system

where the account files are stored.
 A directory is like a Windows folder (more on this
later).
 Many unix commands and applications make use
of the account home directory (as a place to look
for customization files).
 Shell

 A Shell is a unix program that provides an

interactive session - a text-based user interface.
 When you log in to a Unix system the program
you initially interact with is your shell.
 There are a number of popular shells that are
available.
 Popular Shells

sh Bourne Shell
ksh Korn Shell
csh C Shell
bash Bourne-Again Shell
 Startup files
sh,ksh:
/etc/profile (system defaults)
~/.profile
bash:
~/.bash_profile
~/.bashrc
~/.bash_logout
csh:
~/.cshrc
~/.login
~/.logout
 Additional Password Security
 Later versions of Unix have improved the security for password
encryption as follows:
 Passwords no longer restricted to 8 characters
 Use MD5 instead of DES; gives 128-bit output
 Use “salt”
 Furthermore, the encrypted (hashed) password is removed from
the /etc/passwd file and instead is placed in /etc/shadow
 Restricted access to /etc/shadow – no requirement for it to be world-
readable; only readable by Root
 Much more difficult to launch off-line (dictionary) attack
 /etc/shadow contains additional password information (number of days
before expiry, etc)
 passwd, shadow, group files
unix etc # ls -l passwd shadow group tikai “wheel” grupa
-rw-r--r-- 1 root root 705 Sep 23 15:36 group
-rw-r--r-- 1 root root 1895 Sep 24 18:20 passwd var su uz root;
-rw------- 1 root root 634 Sep 24 18:22 shadow skat /etc/pam.d/
unix etc #

unix root # more /etc/passwd

root:x:0:0:root:/root:/bin/bash
unix root # more /etc/group
bin:x:1:1:bin:/bin:/bin/false
root::0:root
daemon:x:2:2:daemon:/sbin:/bin/false
bin::1:root,bin,daemon
adm:x:3:4:adm:/var/adm:/bin/false
daemon::2:root,bin,daemon
lp:x:4:7:lp:/var/spool/lpd:/bin/false
sys::3:root,bin,adm
sync:x:5:0:sync:/sbin:/bin/sync
adm::4:root,adm,daemon
shutdown:x:6:0:shutdown:/sbin:/sbin/shutdown
tty::5:girtsf
halt:x:7:0:halt:/sbin:/sbin/halt
disk::6:root,adm
...
lp::7:lp
guest:x:405:100:guest:/dev/null:/dev/null
mem::8:
nobody:x:65534:65534:nobody:/:/bin/false
kmem::9:
girtsf:x:1000:100::/home/girtsf:/bin/bash
wheel::10:root,girtsf
dima:x:1001:100::/home/dima:/bin/bash
floppy::11:root
guntis:x:1002:100::/home/guntis:/bin/bash
mail::12:mail
students:x:1003:100::/home/students:/bin/bash
...
unix root #
users::100:games,girtsf
nofiles:x:200:
unix root # more /etc/shadow qmail:x:201:
root:$1$VlYbWsrd$GUs2cptio.rKlGHgAMBzr.:12684:0::::: postfix:x:207:
halt:*:9797:0::::: postdrop:x:208:
... smmsp:x:209:smmsp
guest:*:9797:0::::: slocate::245:
nobody:*:9797:0::::: portage::250:portage
girtsf:$1$u6UEWKT2$w5K28n2iAB2wNWtyPLycP1:12684:0:99999:7::: utmp:x:406:
dima:$1$BQCdIBdV$xzzlj4s8XT6L9cLAmcoV50:12684:0:99999:7::: nogroup::65533:
guntis:$1$fiJF/0BT$Py9JiQQL6icajjQVyMZ7//:12684:0:99999:7::: nobody::65534:
students:$1$wueon8yh$nLpUpNOKr8yTYaEnEK6OJ1:12685:0:99999:7::: unix root #
unix root #
 Users and Ownership: /etc/passwd
 Every File is owned by one of the system’s users – identity is represented by
the user-id (UID)
 Password file assoicate UID with system users.
gates:x:65:20:B. Gates:/home/gates:/bin/ksh

command interpreter
home directory
“real” name
group ID
user ID
[encrypted password]
login name
 /etc/group
 Information about system groups
faculty:x:23:maria,eileen,dkl

list of group members

group ID
[encrypted group password]
group name
 Who is superuser ?

 UID of 0
 Any username can be the superuser.
 Normal security checks and constraints are
ignored for the superuser.
 Superuser is not for casual use.
 Do not login as superuser, use ‘/bin/su’ with “-” option
instead.
 Simple trap to steal superuser

 Premise  Set a trap

 Root’s PATH starts with “.” % cd
 Contents of shell script ‘ls’ % chmod 700 .
#!/bin/sh % touch ./-f
cp /bin/sh ./junk/.ss  To do is just say to
chmod 4555 ./junk/.ss administrator. “I have a
rm –f $0 funny file in my directory I
exec /bin/ls ${1+”$@”} can’t seem to delete.”
 Good root practice

unix root # which ls

/bin/ls

unix root # ls -al `which ls`

-rwxr-xr-x 1 root root 79360 Jul 18 08:03 /bin/ls
unix root #

Do not start root PATH with “.”

 Logging In

 To log in to a Unix machine you can either:

 sit at the console (the computer itself)
 access via the net (using telnet, rsh, ssh, kermit, or
some other remote access client).
 The system prompts you for your username and
password.
 Usernames and passwords are case sensitive!
 Session Startup

 Once you log in, your shell will be started and it

will display a prompt.
 When the shell is started it looks in your home
directory for some customization files.
 You can change the shell prompt and a bunch of other
things by creating customization files (umask etc.)
 Your Home Directory

 Every Unix process* has a notion of the “current working

directory”.
 You shell (which is a process) starts with the current
working directory set to your home directory.

*A process is an instance of a program that is

currently running.
 Interacting with the Shell

 The shell prints a prompt and waits for you to type

in a command.
 The shell can deal with a couple of types of
commands:
 shell internals - commands that the shell handles
directly.
 External programs - the shell runs a program for you.
 File Types In Unix

All Files

Text: Readable
Binary: Uses all
characters
characters

Documents, etc. Directories

Source: Readable
Programs
Programming
Compiler
Language: Machine Code:
Shell scripts: Interpreted or Directly executed
Interpreted by Compiled
shell

Executable
Files
to new files
 umask: Calculations (2)
 If you want a file permission of 644 (by default, without
manually executing chmod) on a regular file, the umask
would need to be 022.
Default Mode 666
umask -022
New File Mode 644
 Bit level: new_mask = mode & ~umask
umask = 000010010 = ---rw-rw = 0022
~umask = 111101101
mode = 110110110 = rw-rw-rw = 0666
new_mask = 111100100 = rw------ = 0600
 Advanced: Capabilities
For the purpose of performing permission checks, traditional Unix
implementations distinguish two categories of processes: privi-
leged processes (whose effective user ID is 0, referred to as
superuser or root), and unprivileged processes (whose effective
UID is non-zero). Privileged processes bypass all kernel permis-
sion checks, while unprivileged processes are subject to full per-
mission checking based on the process's credentials (usually:
effective UID, effective GID, and supplementary group list).

Starting with kernel 2.2, Linux provides an (as yet incomplete)

system of capabilities, which divide the privileges traditionally
associated with superuser into distinct units that can be indepen-
dently enabled and disabled.
 Advanced: Access Control Lists
The permissions defined by ACLs are a superset of the permissions specified by the file
permission bits. The permissions defined for the file owner correspond to the permissions
of the ACL_USER_OBJ entry. The permissions defined for the file group correspond to
the permissions of the ACL_GROUP_OBJ entry, if the ACL has no ACL_MASK entry. If
the ACL has an ACL_MASK entry, then the permissions defined for the file group
correspond to the permissions of the ACL_MASK entry. The permissions defined for the
other class correspond to the permissions of the ACL_OTHER_OBJ entry.

Modification of the file permission bits results in the modification of the permissions in
the associated ACL entries. Modification of the permissions in the ACL entries results in
the modification of the file permission bits.

Example:
user::rw-
user:lisa:rw-
group::r--
group:toolies:rw-
mask::r--
other::r--
Advanced: TCP/IP Firewalls