Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Thermoptic Camouflage: Total IDS Evasion

HD Moore
August 03, 2006

Thermoptic Camouflage: Total IDS Evasion

This presentation describes a number of ways to bypass intrusion detection systems and was originally presented at Black Hat USA 2006. This was co-presented by HD Moore and Brian Caswell.

HD Moore

August 03, 2006
Tweet

More Decks by HD Moore

Other Decks in Technology

Transcript

  1. 2 Who are we • Brian Caswell – Principal Research

    Engineer, Sourcefire – Metasploit Developer – The Shmoo Group • H D Moore – Director of Security Research, BreakingPoint Systems – Metasploit Founder
  2. 3 Our Plan for Today • Evasion at every layer

    • Attacking the IDS • Demonstration
  3. 4 Evasion Principles • Know your target – Abuse target-specific

    behavior • Know your network – Abuse TTL and routing issues • Know your IDS – Abuse signature matching engines – Abuse hardware limitations
  4. 5 Evasion Layers • Hardware – Layers 1-2 • Operating

    System – Layers 3-4 • Application – Layers 5-7
  5. 6 Driver Modeling - Evasion at Layer 2 Ethernet ambiguities

    • Slightly oversized frames • Broadcast destinations • Multiple VLAN headers Not useful or practical • Requires local media access • IPS likely to drop the frame
  6. 7 OS Modeling - Evasion at Layer 3/4 Ptacek &

    Newsham • TTL tricks • IP fragmentation • TCP fragmentation • TCP sequence issues Other tricks • Host vs network filtering • Fake connection tear-downs
  7. 8 Fragmentation - Overview IP fragmentation for newbies • Split

    an IP packet into fragments • Minimum fragment size is 8 bytes IP stacks handle this different ways • Overlaps, duplicates, gaps, oh my! • Abuse differences to evade IDS
  8. 9 Fragmentation Models • Paxson & Handley • BSD •

    BSD-Right 1 1 1 4 4 4 4 2 2 3 3 5 5 5 6 6 6 3 1 1 1 4 4 2 3 3 6 6 6 3 1 1 1 4 4 2 3 3 6 6 6 3 4 2 4
  9. 10 Fragmentation Models • Linux • First (Windows) • Last

    (IOS) 1 1 1 4 4 4 4 2 2 3 3 5 5 5 6 6 6 3 1 1 1 2 4 4 5 5 5 6 6 6 1 1 1 4 2 2 3 3 6 6 6 3 1 4 4 4 4 2 5 5 5 6 6 6
  10. 11 Fragmentation - Complications • Novak/Sturges Model • NONE -

    Drop frags (New IOS) • Fragmentation - BSD-Right 1 1 1 4 4 4 4 2 2 3 3 5 5 5 6 6 6 3 7 7 1 4 4 4 2 2 5 5 5 6 7 7
  11. 12 Fragmentation - Windows/Solaris • Windows / Solaris 1 1

    1 4 4 4 4 2 2 3 5 5 5 3 5 1 1 1 4 2 2 5 5 5 5
  12. 13 Application Modeling - Layer 5/6 Millions of applications •

    Protocol code differences • Vendor specific extensions • Error condition handling Fun tricks for every protocol! • HTTP, FTP, SMTP, DNS • SunRPC, DCERPC, SMB
  13. 14 Application Modeling - Layer 5/6 Sidestep, Robert Graham of

    ISS • SunRPC fragmentation • FTP telnet option negotiation • HTTP URI encoding Many new techniques • Vendor-specific URI encoding • Evasion for 'deep' protocols • Client-side attack evasion
  14. 15 SMB Evasions SMB is a transport protocol • Remote

    file access • System administration • Network authentication • Remote procedure calls
  15. 16 SMB Evasions SMB based vulnerabilities • Malware propagation •

    Remote registry access • Authentication attacks • DCERPC transport – MS04-011, MS04-007 – MS05-039, MS06-025 – Distributed COM
  16. 17 SMB Evasions What is an IDS to do? •

    Signature-only • State track + signature • State + context + signature • Complete protocol emulation – What version of the protocol? – What version-specific options? – What vendor-specific options?
  17. 18 SMB Evasions (before & after) Segmented read and write

    operations • Independent of TCP and IP layers • IDS must track length and offset • Evade DCERPC signatures • Evade malware signatures • Offset value ignored for pipes Demonstration
  18. 19 SMB Evasions (before & after) Data and parameter padding

    • Offset value determines location • Used to align fields in memory • Abused to fill with bogus data • Max padding is ~4000 bytes... – Insert fake SMB requests – Trigger low-risk signatures Demonstration
  19. 20 SMB Evasions (before & after) The SMB ECHO command

    • Simple command echos data • Useful to trick SMB state engines • Max size is greater than MTU... – Insert fake SMB responses – Trigger low-risk signatures
  20. 21 SMB Evasions (before & after) SMB Transaction “PIPE” string

    • Normally just \PIPE\ • Not validated by the OS • Max length is ~4000 bytes • Evade almost all Trans signatures • Also useful for state engine attacks Demonstration
  21. 22 SMB Evasions (before & after) SMB CreateAndX Path Names

    • Paths are normalized by target • Trivial to obfuscate with \\\\\\\\\\\\\\ • Evade many DCERPC signatures • Evade malware signatures Demonstration
  22. 23 SMB Evasions (before & after) Unicode & Non-Unicode Strings

    • Evade signatures with Unicode off • All Unicode-based evasions apply • Remember the IIS Unicode bug? • Same thing applies to SMB paths Demonstration
  23. 24 SMB Evasions (before & after) Evasion with AndX Chains

    • Multiple commands per request • Login, open file, write, close, logoff • Also useful for state engine attacks Evasion with Request Stacking • Pile all SMBs into one send() • Side effect of Nagle algorithm
  24. 25 DCERPC Evasion Lets talk DCERPC • Supports multiple transports

    – TCP, HTTP, UDP, SMB (4+ ways) • Multiple ways to represent data – Big endian or little endian byte order – Unicode, EBCDIC, or ASCII strings – IEEE, VAX, Cray, or IBM floats • Supports fragmentation – IP + TCP + SMB + DCERPC = ? • Excellent source of new vulns :-)
  25. 26 DCERPC Evasion DCERPC Basics • Connect to the transport

    • Bind to specific UUID and version • Call function by number Function parameters • Encoding specified by client • Uses the NDR encoding system
  26. 27 DCERPC Evasion DCERPC Bind evasions • Bind to multiple

    UUIDs at once • Bind to one UUID then AlterContext • Bind with authentication field
  27. 28 DCERPC Evasion DCERPC Call evasions • Fragment data across

    many requests • Encrypt data with packet privacy • Append random data to NDR stub • Prepend an Object ID
  28. 29 DCERPC Evasion DCERPC Transport evasion • RPC over HTTP

    via RpcProxy • One-packet UDP function calls – Uses the idempotent flag Ports and processes • Shared processes share pipes • Choose which named pipe to use • Everyone loves \BROWSER
  29. 30 DCERPC - NDR Strings “ABCDE” in Little Endian ASCII

    • Len + Offset + TotalLen + string + null pad to 32bit boundary "\x05\x00\x00\x00” "\x00\x00\x00\x00” "\x05\x00\x00\x00” "ABCD” "E\x00\x00\x00" • Use non-NULLs for padding!
  30. 31 DCERPC - NDR Strings Empty string "" in Little

    Endian ASCII • Len + Offset + TotalLen + string + pad to 32bit boundary "\x00\x00\x00\x00” "\x00\x00\x00\x00” "\x00\x00\x00\x00” "\x00\x00\x00\x00” • Or on some services "\x00\x00\x00\x00"
  31. 32 DCERPC - ISystemActivator Blaster NDR stub *(DWORD *)(buf2+0x10)=*(DWORD *)(buf2+0x10)+len-0xc;

    *(DWORD *)(buf2+0x80)=*(DWORD *)(buf2+0x80)+len-0xc; *(DWORD *)(buf2+0x84)=*(DWORD *)(buf2+0x84)+len-0xc; *(DWORD *)(buf2+0xb4)=*(DWORD *)(buf2+0xb4)+len-0xc; *(DWORD *)(buf2+0xb8)=*(DWORD *)(buf2+0xb8)+len-0xc; *(DWORD *)(buf2+0xd0)=*(DWORD *)(buf2+0xd0)+len-0xc; *(DWORD *)(buf2+0x18c)=*(DWORD *)(buf2+0x18c)+len-0xc;
  32. 33 DCERPC - ISystemActivator How did vendors look for this

    attack? • MEOW prefixes for objects 4d45 4f57 0400 0000 a201 0000 0000 0000 MEOW............ c000 0000 0000 0046 3803 0000 0000 0000 .......F8....... c000 0000 0000 0046 0000 0000 f005 0000 .......F........ e805 0000 0000 0000 0110 0800 cccc cccc ................ c800 0000 4d45 4f57 e805 0000 d800 0000 ....MEOW........ • Long Paths – \\TOO_LONG_PATH_HERE\
  33. 34 DCERPC - ISystemActivator Path Contains 8 objects, bad one

    is #7 • Paths everywhere! • One object allows ~1Mb of padding! • All Windows path rules also apply
  34. 35 Text Protocols: Header Folding Header parsing is ambiguous •

    HTTP, SMTP, iCal, Email • EvilHeader: Bar Biz; boo • What does your application do? – "EvilHeader: Bar Bi\n ;boo" – "EvilHeader: Bar Biz\n boo"
  35. 36 Client-Side Attack Evasion Juicy targets for many reasons –

    No firewall, rich text, scripting, many bugs So many evasion options • Unicode • Javascript • Objects • Compression • Encryption
  36. 37 Unicode Which Unicode? • utf-16le, utf-16be, utf32-le, utf32-be, •

    utf-7, utf-8 • HTTP – Content-Type: text/html; charset: utf-16be • Oops – Start with "\xFE\xFF", forces utf-16be
  37. 38 UTF-8 Overlong Strings • Encode the letter “A” –

    41, c1a1, e081a1, f08081a1, f8808081a1, fc80808081a1 • “Invalid” overlong strings – 2 bytes • c121, c161, c1a1, c1e1 – 3 bytes • e00101, e00141, e00181, e001c1, e04101, e04141, e04181, e041c1, e08101, e08141, e08181, e081c1, e0c101, e0c141, e0c181, e0c1c1 • 125 ways in ONE character set!
  38. 39 Common Javascript Evasions <script>document.write(”EVIL") </script> • <body onLoad= ”document.write('EVIL')”;>

    • document.write( unescape( '%45%56%49%4C')); • <font style=“background: url(javascript:document.write(‘EVIL));”>
  39. 40 Uncommon Javascript Evasions • <font style=“background: url(jav as c

    ript:alert(‘evil’));”> • <scr\xFE\xFFipt> alert('CVE-2006-2783'); </s\xFE\xFFript> • Using PCRE to strip javascript? • Unicode (default doesn’t support it…) • Rejects overlong strings, 0xFF, or 0xFE
  40. 41 Base64 your HTML <OBJECT ID="w00t" TYPE="text/html" DATA="data:text/html;base64,ZXZpb CB0ZXh0">aww, too

    bad!</OBJECT> • Equivalent to "evil text” • Don't write signatures for this! • Spaces matter – "foo”, ” foo" , " foo” – IGZvbw==, ICBmb28=, ICAgZm9v
  41. 42 Compression Issues • Zip Bombs – 100Mb => 100k

    - GZIP • Who writes rules for GZIP output? – WMF header • Arbitrary sized headers in GZIP – name, comment • Three compression algorithms – gzip, deflate, compress
  42. 43 SSL your attacks Encryption is fun • Purchase a

    certificate ($$$) • Compromise and hijack existing cert • Convince the user to ignore warnings • Use SSL wrapped CGI proxy server! https://www.fsurf.com/index.php?q=http://IP:8080/foo.pls https://proxify.com/u?http://IP:8080/foo.wmf
  43. 44 Attacking the IDS Find the failure points • Alert

    management • Hardware limitations • Session tracking • Pattern matching • Signature strength
  44. 45 IDS Alert Management Attack the software • Flood the

    alert system • Nikto is great for this! • Multiple alerts per packet? – One IDS triggers ~1050 per packet! Attack the user • Hide the real attack in the flood • Abuse UI limitations to hide events
  45. 46 IDS Hardware Limitations Gigabit Ethernet limits • 1,000,000,000 bits

    • 125000000 bytes • 1602564 packets • 1.602 packets per microsecond • Oh, full duplex... • 3.205 packets per microsecond
  46. 47 IDS Hardware Limitations PC hardware limitations • PCI/PCI-X limits

    – 33Mhz: 32/64 = 133/266 MB/s – 66Mhz: 32/64 = 266/532 MB/s – 100Mhz: 64 = 800MB/s • Software interrupt limits – Intel Pro/1000 Server / 3Ghz P4/Xeon – 680,000pps RX | 840,000pps TX – 348Mbps capture w/64b packets * Poll mode bypasses interrupt limits
  47. 48 IDS Hardware Limitations PC hardware realities • Typical Dell

    1U appliance – Dual Intel Pro/1000 cards – 3.0Ghz Xeon • 760Mbps max capture mode • 380Mbps max inline mode • The ICSA report agrees! – ISS G400 Proventia rated at 350Mbps
  48. 49 IDS Hardware Limitations Network hardware realities • FastPath vs

    SlowPath – Minimum processing on FastPath – SlowPath used for exceptions • Find the SlowPath – Management services – Encryption and authentication – IP fragment processing
  49. 50 IDS Hardware Limitations Shared cores for hardware • A

    “core” is licensed for a chip – Provides common networking features – Routing, reassembly, switching, etc • Quickest way to add a feature – Common choice for quick development – Just as buggy as any other software – Any flaw applies to multiple vendors :-)
  50. 51 IDS Hardware Limitations Memory allocation • Static blocks preferred

    over allocator • Block must hold entire packet • Split into “buckets” based on size • Stream a specific packet pattern – Try 63, 65, 129, 257, 1025, 2049 – Allocate all blocks in a given bucket – Force exceptions and pass-through
  51. 52 Session Tracking Limitations Hash Collisions • Crosby & Wallach

    • key = srcip ^ dstport • 2^16 srcip/dstports hash equally • data[key] -> Linked List ip/port • Force walking the linked list • 43.78 minutes for 65k PACKETS
  52. 53 Session Tracking Limitations Splay Trees • Self-balancing binary tree

    • O(log(n)) amortized over time • Worst Case = Sorted List • O(n) to rebalance from worst case Demonstration
  53. 54 Attacking Pattern Matchers • Find the most expensive operation

    – Force it to repeat over and over • Trigger exception processing – Use invalid characters, recursion, etc • Inject termination characters – Use terminator strings to fail a match – Depends on the signature and protocol
  54. 55 Attacking Pattern Matchers char * search(char *buf, int buflen,

    char *string, int stringlen) { char *ptr = buf; int i = 0; while ( (i + stringlen) < buflen ) { if ( memcmp(ptr, string, stringlen) == 0 ) { return ptr; } i++; ptr++; } return NULL; }
  55. 56 Attacking Pattern Matchers search(data, datalen, "evilfoo!", 8); • Maximize

    work done by memcpy • Send “evilfoo” * 8 • 48 calls to memcpy • 96 to 384 memory operations0 • 2000 ms on a 65k packet of evilfoo [0] Depending on platform, alignment, and libc implementation
  56. 57 Attacking Pattern Matchers /.*From=[^&]{165,}.*/ • .* – Match any

    amount of any character • From= • [^&]{165,} – 165 or more bytes of anything but & • Force repeated backtrack – “From=“ repeating, “&” at byte 165 Demonstration
  57. 58 Attacking the Signatures • Difference between IDS and application

    – isspace • \t, \n, \v, \f, \r or “ “ – Newlines • \r, \n, \r\n • Force signature engine to stop early – Hit memory limits • PCRE_CONFIG_POSIX_MALLOC_THRESHOLD – Hit recursion limits • PCRE_CONFIG_STACKRECURSE – Hit maxiumum failure limits • PCRE_CONFIG_MATCH_LIMIT
  58. 59 Extracting signatures • Blackbox signature discovery – Create protocol

    template, set boundaries – Enable block mode in IPS product – Flood request permutations and create sig :-) • Direct memory access – Hardware bus monitoring – Root the box and dump the process • Poor cryptography – Key has to accessible somewhere
  59. 60 Conclusion Everything can be evaded • At what layer?

    • At what cost? • At what speed?
  60. 61 Contact Brian Caswell • bmc[at]shmoo.com • http://www.shmoo.com/~bmc/ H D

    Moore • hdm[at]metasploit.com • http://metasploit.com/