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

High Speed Bug Discovery with Fuzzing (version with presenter notes) - Dog Food Conference

High Speed Bug Discovery with Fuzzing (version with presenter notes) - Dog Food Conference

How can you find hidden defects in your code without a lot of manual analysis? You’ll leave this talk knowing when to use fuzzing to test your application, which tools you should use, how to implement a fuzzer from scratch, and when other techniques are a better choice.

Craig Stuntz

October 05, 2017

More Decks by Craig Stuntz

Other Decks in Programming


  1. How Do We Design Software Security? - Usability Studies (Does

    it work for human beings?) - Threat Modeling - Formal Verification - Static Analysis - Auditing - Pentesting - Throw Random Crap At It And See If It Dies
  2. “ Unfortunately, C and C++ are mostly taught the old

    way, as if programming in them isn’t like walking in a minefield. Nor have the books about C and C++ caught up with the current reality. These things must change. -John Regehr https://blog.regehr.org/archives/1520 Fuzzing is depressing. It Shouldn’t Be Necessary, C (and infrastructure to sustain it), Better Methods of Verification Exist, Can Miss Stuff , Worse Is Better? Sometimes when something seems weird to you it’s worth taking a deep dive,
  3. “Yet, despite the crippling and obvious limitations of fuzzing and

    the virtues of symbolic execution, there is one jarring discord: I'm fairly certain that probably around 70% of all remote code execution vulnerabilities disclosed in the past few years trace back to fairly “dumb” fuzzing tools, with the pattern showing little change over time. -Michał Zalewski https://lcamtuf.blogspot.com/2015/02/symbolic-execution-in-vuln-research.html
  4. let private bitMasks(bitIndex: int, bitsToFlip: int) = match bitIndex %

    8, bitsToFlip with | 0, 1 !→ 0b00000001uy, 0b00000000uy | 1, 1 !→ 0b00000010uy, 0b00000000uy | 2, 1 !→ 0b00000100uy, 0b00000000uy | 3, 1 !→ 0b00001000uy, 0b00000000uy | 4, 1 !→ 0b00010000uy, 0b00000000uy | 5, 1 !→ 0b00100000uy, 0b00000000uy | 6, 1 !→ 0b01000000uy, 0b00000000uy | 7, 1 !→ 0b10000000uy, 0b00000000uy | 0, 2 !→ 0b00000011uy, 0b00000000uy | 1, 2 !→ 0b00000110uy, 0b00000000uy | 2, 2 !→ 0b00001100uy, 0b00000000uy | 3, 2 !→ 0b00011000uy, 0b00000000uy | 4, 2 !→ 0b00110000uy, 0b00000000uy | 5, 2 !→ 0b01100000uy, 0b00000000uy | 6, 2 !→ 0b11000000uy, 0b00000000uy | 7, 2 !→ 0b10000000uy, 0b00000001uy | 0, 4 !→ 0b00001111uy, 0b00000000uy | 1, 4 !→ 0b00011110uy, 0b00000000uy | 2, 4 !→ 0b00111100uy, 0b00000000uy | 3, 4 !→ 0b01111000uy, 0b00000000uy | 4, 4 !→ 0b11110000uy, 0b00000000uy | 5, 4 !→ 0b11100000uy, 0b00000001uy | 6, 4 !→ 0b11000000uy, 0b00000011uy | 7, 4 !→ 0b10000000uy, 0b00000111uy | bit, _ !→ failwithf "Unsupported bit %d or bitsToFlip %d" bit bitsToFlip so I wrote some really terrible and embarrassing F# to learn about fuzzing. A lot of it is just literally translated from C.
  5. Spoilers! Why should I care? (because it’s surprisingly effective at

    finding bugs in software) What is it? (a simple, property-based randomized testing technique) When should I use it? (integration testing complex systems with infinite input values) How do I get started? (I’ll suggest a bunch of tools) Should I write my own? (yes, and I have stories!) When should I not use it? Will start very non-technical and get more technical as we go on. Light bash scripting to rewriting binary assemblies
  6. 400 Crashes, 106 Distinct Security Bugs in Adobe Flash Player

    https://security.googleblog.com/2011/08/fuzzing-at-scale.html 1 month, 2000 cores
  7. 325 C Compiler bugs in GCC, Clang, & Others https://www.flux.utah.edu/paper/yang-pldi11

    25 GCC bugs classified as release-blocking Spent less than $1000
  8. “fuzzers…can be integrated into a CI environment for permanent protection

    henceforth” Guido Vranken https://guidovranken.wordpress.com/2017/06/21/the-openvpn-post-audit-bug-bonanza/ Audits are one-time, fuzzers run forever
  9. Prevent Regressions Bug Discovery Help with Code Design Meets Specifications

    Fuzzing Integration testing Unit testing Formal verification Exploratory testing Testing goals and tools
  10. How Many Cases Should We Test? One Only the Most

    Interesting Every Possible Case Unit Testing Fuzzing Formal Verification Formal verification example: a type signature
  11. Corpus A few handwritten examples Fuzzing databases Harvest from test

    suites, defect reports Harvest from public Internet
  12. System Under Test A function Entire application Part of OS

    kernel Many testing techniques limited to testing at a particular scope
  13. Properties Does it crash? Does it hang? Is the output

    “valid”? Does execution trip an address or memory sanitizer? Does the output match some other system?
  14. Magic Mutation of corpus Coverage guidance Lots of test runs

    What you’ve heard so far sounds like taking a few sample inputs and doing a boring test on the output. That’s it?
  15. Possible Inputs Random Inputs Interesting Inputs Random Inputs with Profile

    Guidance The idea is to find the interesting (desired properties don’t hold) inputs to the system under test faster than we could find them by either exhaustive testing or purely by chance.
  16. Getting Started with afl - Compile system under test with

    instrumentation - Place corpus input(s) in a folder - Invoke afl - Wait for bugs https://fuzzing-project.org/tutorial3.html What does this look like in practice? There are more complicated options, but this is how you get started Sounds kinda dumb but scales really well
  17. Place Corpus in Folder $ mkdir in $ cd in

    $ cat > foo.json { "a": "bc" } ^D $ cd .. You probably want more than one example in the corpus, but this gets us started
  18. Invoke afl $ afl-fuzz -i in -o out \ my_json_parser

    @@ folder containing corpus “@@“ means “the current test case” system under test findings go here This is the simple case. There are many options!
  19. This might run for weeks What will I find? Crashes,

    hangs Can do more, with effort!
  20. afl In a Nutshell ⃗ Turn a few maybe uninteresting

    inputs into a lot of hopefully very interesting inputs, quickly
  21. afl Fuzz Strategies Walking bit flips (try flipping each bit

    in input individually) Walking byte flips (try flipping each contiguous set of 8 bits) Simple arithmetic (increment or decrement bytes in the file by certain small values) Known integers (or dictionaries) (replace bytes with “problematic” 8, 16, and 32 bit integers like 0 and FF) Profile-guided stacked tweaks and test case splicing (magic!) https://lcamtuf.blogspot.com/2014/08/binary-fuzzing-strategies-what-works.html Deterministic, genetic
  22. Walking Bit Flip Original 01010101 Flip bit 0 01010100 Flip

    bit 1 01010111 Flip bit 2 01010001 <etc.> Walking 2 Bit Flip Original 01010101 Flip bits 0,1 01010110 Flip bits 2,1 01010011 Flip bits 3,2 01001101 <etc.>
  23. Unit Tests Fuzzing Useful For Preventing Regressions, Design Finding New

    Bugs Tests Functions Any Level Test Examples Hand-selected values Corpus + Mutation Execution Time Milliseconds Weeks Magic? No Yes
  24. $ ./configure CC="afl-gcc" \ CXX="afl-g++" \ --disable-shared; \ make You’ve

    seen this before; afl requires recompiling with instrumentation
  25. STJSON A JSON Parser in Swift 3 compliant with RFC

    7159 STJSON was written along with the article Parsing JSON is a Minefield. Basic usage: var p = STJSONParser(data: data) do { let o = p.parse() } catch let e { print(e) } Instantiation with options: var p = STJSON(data:data, maxParserDepth:1024, options:[.useUnicodeReplacementCharacter]) https://github.com/nst/STJSON https://github.com/CraigStuntz/Fizil/tree/master/StJson Is there a specification for the behavior of the program?
  26. Dumb Fuzzer Mangling Byte Arrays many interesting programs take binary

    input sometimes this requires less domain knowledge (not always!) or at least less custom code What about programs or functions which can’t accept binary input?
  27. public boolean SomeFunction( SomeEnum firstArg, int secondArg) { ✗ not

    a good candidate for dumb fuzzing If I have a good property I can test exhaustively
  28. Smart Fuzzing MongoDB Expression Grammar http://queue.acm.org/detail.cfm?ref=rss&id=3059007 Not an end to

    end test — testing the internals. Don’t test stuff which takes forever
  29. Isn’t it just for Security? https://www.flickr.com/photos/wocintechchat/25721078480/ Every security person knows

    about fuzzing. Almost no app devs do. Security people care more about bugs than app devs.
  30. How to Get Started with Fuzzing 1. Find a program

    to test 2. Find a fuzzer 3. Find a corpus 4. Choose a property 5. Let it run!
  31. $ gzip -c /bin/bash > sample.gz $ while true do

    radamsa sample.gz > fuzzed.gz gzip -dc fuzzed.gz > /dev/null test $? -gt 127 && break done ← Fuzz the corpus ← Execute S.O.T. ← Check a property ← Repeat a lot! https://github.com/aoh/radamsa Radamsa: Un*x philosophy of “do one thing well, chain with other stuff” Works on Windows with cygwin (and maybe
  32. afl

  33. OSS-Fuzz Submit your project https://github.com/google/oss-fuzz 1. You can use a

    fuzzer in your CI 2. Since it’s easy to do so, Google did it for you.
  34. “ We didn't call it fuzzing back in the 1950s,

    but it was our standard practice to test programs by inputting decks of punch cards taken from the trash. -Gerald M. Weinberg http://secretsofconsulting.blogspot.com/2017/02/fuzz-testing-and-fuzz-history.html
  35. Fuzzing SQLite with afl Start with a single test case:

    create table t1(one smallint); insert into t1 values(1); select * from t1; Add a list of reserved words from documentation Then extract SQL statements from SQLite unit tests (550 files at around 220 bytes each) https://lcamtuf.blogspot.com/2015/04/finding-bugs-in-sqlite-easy-way.html
  36. One of The Things …Is Not Like the Others! Repeated

    testing with same input should yield same output
  37. Interesting Stuff I Learned While Writing a Fuzzer - Deceptive

    simplicity of fuzzing - F# bitwise operations - How to instrument .NET code - dnSpy is awesome - Same input -> Same code -> Different paths - Strong naming is painful - Unicode is also painful - Speed is everything
  38. let jsonNetResult = try JsonConvert.DeserializeObject<obj>(str) |> ignore Success with |

    :? JsonReaderException as jre -> jre.Message |> Error | :? JsonSerializationException as jse -> jse.Message |> Error | :? System.FormatException as fe -> if fe.Message.StartsWith("Invalid hex character”) // hard coded in Json.NET then fe.Message |> Error else reraise() ⃪ T est ⬑ Special case error stuff Special test harness code specific to the system under test
  39. use proc = new Process() proc.StartInfo.FileName <- executablePath inputMethod.BeforeStart proc

    testCase.Data proc.StartInfo.UseShellExecute <- false proc.StartInfo.RedirectStandardOutput <- true proc.StartInfo.RedirectStandardError <- true proc.StartInfo.EnvironmentVariables.Add(SharedMemory.environmentVariableName, sharedMemoryName) let output = new System.Text.StringBuilder() let err = new System.Text.StringBuilder() proc.OutputDataReceived.Add(fun args -> output.Append(args.Data) |> ignore) proc.ErrorDataReceived.Add (fun args -> err.Append(args.Data) |> ignore) proc.Start() |> ignore inputMethod.AfterStart proc testCase.Data proc.BeginOutputReadLine() proc.BeginErrorReadLine() proc.WaitForExit() let exitCode = proc.ExitCode let crashed = exitCode = WinApi.ClrUnhandledExceptionCode ⃪ Set up ⃪ Read results ⃪ Important bit Generic code in Fizil
  40. /// An ordered list of functions to use when starting

    with a single piece of /// example data and producing new examples to try let private allStrategies(dictionaryValues: byte[][])= [ bitFlip 1 bitFlip 2 bitFlip 4 byteFlip 1 byteFlip 2 byteFlip 4 arith8 arith16 arith32 dictionary dictionaryValues interest8 interest16 ]
  41. let totalBits = bytes.Length * 8 let testCases = seq

    { for bit = 0 to totalBits - flipBits do let newBytes = Array.copy bytes let firstByte = bit / 8 let firstByteMask, secondByteMask = bitMasks(bit, flipBits) let newFirstByte = bytes.[firstByte] ^^^ firstByteMask newBytes.[firstByte] <- newFirstByte let secondByte = firstByte + 1 if secondByteMask <> 0uy && secondByte < bytes.Length then let newSecondByte = bytes.[secondByte] ^^^ secondByteMask newBytes.[secondByte] <- newSecondByte yield newBytes } Fuzz one byte → ^^^ means xor ↓
  42. private static void F(string arg) { #if MANUAL_INSTRUMENTATION instrument.Trace(29875); #endif

    Console.WriteLine("f"); Console.Error.WriteLine("Error!"); Environment.Exit(1); }
  43. let stringify (ob: obj) : string = JsonConvert.SerializeObject(ob) // Method:

    System.String\u0020Program::stringify(System.Object) .body stringify { arg_02_0 [generated] arg_07_0 [generated] nop() arg_02_0 = ldloc(ob) arg_07_0 = call(JsonConvert::SerializeObject, arg_02_0) ret(arg_07_0) } // Method: System.String\u0020Program::stringify(System.Object) .body stringify { arg_05_0 [generated] arg_0C_0 [generated] arg_11_0 [generated] arg_05_0 = ldc.i4(23831) call(Instrument::Trace, arg_05_0) nop() arg_0C_0 = ldloc(ob) arg_11_0 = call(JsonConvert::SerializeObject, arg_0C_0) ret(arg_11_0) }
  44. let private insertTraceInstruction(ilProcessor: ILProcessor, before: Instruction, state) = let compileTimeRandom

    = state.Random.Next(0, UInt16.MaxValue |> Convert.ToInt32) let ldArg = ilProcessor.Create(OpCodes.Ldc_I4, compileTimeRandom) let callTrace = ilProcessor.Create(OpCodes.Call, state.Trace) ilProcessor.InsertBefore(before, ldArg) ilProcessor.InsertAfter (ldArg, callTrace) This margin is too narrow to contain a try/finally example, so see: https://goo.gl/W4y7JH Inserting the IL instructions I needed was fairly easy. Here is the important bit of the code which does it. How did I learn how to write this? I instrumented a small program “manually” by writing the instrumentation code myself, and then decompiled that program to figure out which IL instructions I needed. Inserting them with Mono.Cecil is just a few lines of code. try/finally is much, much harder. I won’t even try to walk you through it here. Look at the GitHub repo if you want to see how it’s done.
  45. http://www.json.org/ I need a way to determine if Json.NET is

    parsing the JSON correctly. So I thought I should write a JSON validator to check its behavior. Fortunately, there’s a standard! “Probably the boldest design decision I made was to not put a version number on JSON so there is no mechanism for revising it. We are stuck with JSON: whatever it is in its current form, that’s it.” -Crockford
  46. https://tools.ietf.org/html/rfc7159 And another JSON standard. And no, they don’t all

    agree on everything, nor is there a single, “latest” version. Despite this multitude of standards, there are still edge cases intentionally delegated to the implementer — what we would call “undefined behavior” in C.
  47. https://github.com/nst/STJSON I was going to write my own validator, but…

    Nicolas Seriot wrote a validator called STJSON which attempts to synthesize these as much as possible.
  48. https://github.com/CraigStuntz/Fizil/blob/master/StJson/StJsonParser.fs Swift doesn’t readily compile to Windows, but if you

    squint hard enough it kind of looks like F#, so I ported the code and used it to validate Json.NET's behavior.
  49. Standard Rejects, Json.NET Accepts Value [,,,] Standard Says A JSON

    value MUST be an object, array, number, or string, or one of the following three literal names: false null true Json.NET [null, null, null, null] Things JSON.NET succeeds on that the standard rejects
  50. Strong naming was a consistent pain for me. I’m altering

    the binaries of assemblies, and part of the point of strong naming is to stop you from doing just that, so naturally if the assembly is strongly named it can’t be loaded when I’m finished.
  51. let private removeStrongName (assemblyDefinition : AssemblyDefinition) = let name =

    assemblyDefinition.Name; name.HasPublicKey <- false; name.PublicKey <- Array.empty; assemblyDefinition.Modules |> Seq.iter ( fun moduleDefinition -> moduleDefinition.Attributes <- moduleDefinition.Attributes &&& ~~~ModuleAttributes.StrongNameSigned) let aptca = assemblyDefinition.CustomAttributes.FirstOrDefault( fun attr -> attr.AttributeType.FullName = typeof<System.Security.AllowPartiallyTrustedCallersAttribute>.FullName) assemblyDefinition.CustomAttributes.Remove aptca |> ignore assembly.MainModule.AssemblyReferences |> Seq.filter (fun reference -> Set.contains reference.Name assembliesToInstrument) |> Seq.iter (fun reference -> reference.PublicKeyToken <- null ) So I need to remove the strong name from any assembly I fuzz, but I also need to remove the PublicKeyToken from any other assembly which references it. Doing this in Mono.Cecil is not well-documented, and after quite a bit of time spent in GitHub issues and trial and error I figured out that it takes 5 distinct steps to do this.
  52. “ “If marked BeforeFieldInit then the type’s initializer method is

    executed at, or sometime before, first access to any static field defined for that type.” -ECMA-335, Common Language Infrastructure (CLI), Partition I
  53. Unicode Original JSON { "a": "bc" } ASCII Bytes 7B

    20 22 61 22 20 3A 20 22 62 63 22 20 7D UTF-8 with Byte Order Mark EF BB BF 7B 20 22 61 22 20 3A 20 22 62 63 22 20 7D UTF-16 BE with BOM FE FF 00 7B 00 20 00 22 00 61 00 22 00 20 00 3A 00 20 00 22 00 62 00 63 00 22 00 20 00 7D
  54. Resources Awesome Fuzzing: A Curated List of Fuzzing Resources https://github.com/secfigo/Awesome-Fuzzing

    OWASP Fuzzing Page https://www.owasp.org/index.php/Fuzzing MongoDB’s JavaScript Fuzzer http://queue.acm.org/detail.cfm?ref=rss&id=3059007 afl technical details http://lcamtuf.coredump.cx/afl/technical_details.txt afl Help Email List [email protected] Fizil https://github.com/CraigStuntz/Fizil WTF, ACM?
  55. Thank You! - Michał Zalewski, for afl documentation - Rehearsal

    audiences, employees of - Dynamit - Improving - Ineffable Solutions