10-15-2018

Last week, I gave a talk at 第51回情報科学若手の会.

I am pleased to be able to tell the relationship between the artisanship of reverse engineering and the principle of computer science. The questions asked at the site are as follows:

Q	A
Does the “operation” here mean mathematical operation or anything else?	In this presentation, the word “operation” corresponds to assembly or BitVector operations.
How much advanced obfuscation is spreading?	Although it is not widely used in malware for masses, it is common in government-grade malware e.g. APT28.

Also, I got the feedback:

Hello - We had done similar work on JEB in 2017 (plugins for our native decompilers) to perform: 1) opaque predicate remove (see https://t.co/pAp5dD25XL) and 2) CFG un-flattening (see https://t.co/3fl5Qpr4g7) – check out the videos.
— JEB Decompiler (@jebdec) 2018年10月8日

It sounds great.

In the 会, I’ve enjoyed discussing various topics related to information science:

Polyhedral optimization, a method to optimize a certain kind of loop called SCoP using a linear algebra model.
The paper about principle of adversarial examples focusing on Fourier basis functions.
Ragel, the parser generator used in the Ruby community. It was news to me.
The hypothesis that the cerebellum may be using reinforcement learning.
The weakness of random number generator of Dragon Quest 4 (PS version).
About quantitative indicator of the beautiness of exploit code.
Some binary analysis platform written in Rust e.g. wasabi, falcon and finch.
The fact that Pokemon Go players are knowledgeable about Hilbert curve.

If these topics capture your interest, why don’t you join the next 会?

picoCTF 2018 Dog or Frog Write-up

10-12-2018

Thank you for telling me this problem, @kumagi!

Introduction

Dog or Frog 2018shell2.picoctf.com:18318 is a classic task about adversarial examples.

The following files are given:

model
solution template
notes
source

In short, what we need to do is adding noise to given image and to lead the classifier to misrecognition.

Solution

I wrote a patch to the solution template with reference to the tutorial and the paper with well-known panda figure.

--- solution_template.py	2018-10-04 03:16:03.024712745 +0900
+++ solve.py	2018-10-04 03:34:32.474374810 +0900
@@ -1,13 +1,17 @@
 from keras.applications.mobilenet import preprocess_input
 from keras.models import load_model
 from keras.preprocessing.image import img_to_array, array_to_img
+from keras.applications.mobilenet import decode_predictions, preprocess_input
 from PIL import Image
 from imagehash import phash
+import numpy as np
+from keras import backend as K
 
 
 IMAGE_DIMS = (224, 224)
 TREE_FROG_IDX = 31
 TREE_FROG_STR = "tree_frog"
+THRESHOLD = 0.95
 
 
 # I'm pretty sure I borrowed this function from somewhere, but cannot remember
@@ -46,7 +50,27 @@
     model = load_model(model_path)
 
     # TODO: YOUR SOLUTION HERE
+    model.summary()
 
+    cost_function = model.layers[-1].output[0, TREE_FROG_IDX]
+    gradient_function = K.gradients(cost_function, model.layers[0].input)[0]
+    grab_cost_and_gradients_from_model = K.function([model.layers[0].input, K.learning_phase()], [cost_function, gradient_function])
+
+    cost = 0.0
+
+    while cost < THRESHOLD:
+        cost, gradients = grab_cost_and_gradients_from_model([test, 0])
+
+        test += 0.007 * np.sign(gradients) 
+
+        print('{}: {}% confidence'.format(TREE_FROG_STR, cost * 100))
+
+    print('_' * 65)
+    preds = model.predict(test)
+    _, label1, conf1 = decode_predictions(preds)[0][0]
+    print('{}: {}% confidence'.format(label1, conf1 * 100))
+    dec_preds = decode_predictions(preds)[0]
+    print(dec_preds)
 
     test = test.reshape((224,224,3))
     img = array_to_img(test)

The meaning of this implementation is the formula below which adjusted to the given model:

$\tilde{\boldsymbol{x}} = \boldsymbol{x} + \epsilon \thinspace \text{sign} (\nabla_\boldsymbol{x} \text{Loss}(\boldsymbol{x}, y))$ where $y$ is a label.

Then I’ve got:

$ python solve.py
Using TensorFlow backend.
2018-10-04 03:16:48.974593: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
/home/ntddk/.pyenv/versions/anaconda3-5.0.0/lib/python3.6/site-packages/keras/engine/saving.py:292: UserWarning: No training configuration found in save file: the model was *not* compiled. Compile it manually.
  warnings.warn('No training configuration found in save file: '
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         (None, 224, 224, 3)       0         
_________________________________________________________________
conv1_pad (ZeroPadding2D)    (None, 225, 225, 3)       0         
_________________________________________________________________
conv1 (Conv2D)               (None, 112, 112, 32)      864       
_________________________________________________________________
conv1_bn (BatchNormalization (None, 112, 112, 32)      128       
_________________________________________________________________
conv1_relu (ReLU)            (None, 112, 112, 32)      0         
_________________________________________________________________
conv_dw_1 (DepthwiseConv2D)  (None, 112, 112, 32)      288       
_________________________________________________________________
conv_dw_1_bn (BatchNormaliza (None, 112, 112, 32)      128       
_________________________________________________________________
conv_dw_1_relu (ReLU)        (None, 112, 112, 32)      0         
_________________________________________________________________
conv_pw_1 (Conv2D)           (None, 112, 112, 64)      2048      
_________________________________________________________________
conv_pw_1_bn (BatchNormaliza (None, 112, 112, 64)      256       
_________________________________________________________________
conv_pw_1_relu (ReLU)        (None, 112, 112, 64)      0         
_________________________________________________________________
conv_pad_2 (ZeroPadding2D)   (None, 113, 113, 64)      0         
_________________________________________________________________
conv_dw_2 (DepthwiseConv2D)  (None, 56, 56, 64)        576       
_________________________________________________________________
conv_dw_2_bn (BatchNormaliza (None, 56, 56, 64)        256       
_________________________________________________________________
conv_dw_2_relu (ReLU)        (None, 56, 56, 64)        0         
_________________________________________________________________
conv_pw_2 (Conv2D)           (None, 56, 56, 128)       8192      
_________________________________________________________________
conv_pw_2_bn (BatchNormaliza (None, 56, 56, 128)       512       
_________________________________________________________________
conv_pw_2_relu (ReLU)        (None, 56, 56, 128)       0         
_________________________________________________________________
conv_dw_3 (DepthwiseConv2D)  (None, 56, 56, 128)       1152      
_________________________________________________________________
conv_dw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       
_________________________________________________________________
conv_dw_3_relu (ReLU)        (None, 56, 56, 128)       0         
_________________________________________________________________
conv_pw_3 (Conv2D)           (None, 56, 56, 128)       16384     
_________________________________________________________________
conv_pw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       
_________________________________________________________________
conv_pw_3_relu (ReLU)        (None, 56, 56, 128)       0         
_________________________________________________________________
conv_pad_4 (ZeroPadding2D)   (None, 57, 57, 128)       0         
_________________________________________________________________
conv_dw_4 (DepthwiseConv2D)  (None, 28, 28, 128)       1152      
_________________________________________________________________
conv_dw_4_bn (BatchNormaliza (None, 28, 28, 128)       512       
_________________________________________________________________
conv_dw_4_relu (ReLU)        (None, 28, 28, 128)       0         
_________________________________________________________________
conv_pw_4 (Conv2D)           (None, 28, 28, 256)       32768     
_________________________________________________________________
conv_pw_4_bn (BatchNormaliza (None, 28, 28, 256)       1024      
_________________________________________________________________
conv_pw_4_relu (ReLU)        (None, 28, 28, 256)       0         
_________________________________________________________________
conv_dw_5 (DepthwiseConv2D)  (None, 28, 28, 256)       2304      
_________________________________________________________________
conv_dw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      
_________________________________________________________________
conv_dw_5_relu (ReLU)        (None, 28, 28, 256)       0         
_________________________________________________________________
conv_pw_5 (Conv2D)           (None, 28, 28, 256)       65536     
_________________________________________________________________
conv_pw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      
_________________________________________________________________
conv_pw_5_relu (ReLU)        (None, 28, 28, 256)       0         
_________________________________________________________________
conv_pad_6 (ZeroPadding2D)   (None, 29, 29, 256)       0         
_________________________________________________________________
conv_dw_6 (DepthwiseConv2D)  (None, 14, 14, 256)       2304      
_________________________________________________________________
conv_dw_6_bn (BatchNormaliza (None, 14, 14, 256)       1024      
_________________________________________________________________
conv_dw_6_relu (ReLU)        (None, 14, 14, 256)       0         
_________________________________________________________________
conv_pw_6 (Conv2D)           (None, 14, 14, 512)       131072    
_________________________________________________________________
conv_pw_6_bn (BatchNormaliza (None, 14, 14, 512)       2048      
_________________________________________________________________
conv_pw_6_relu (ReLU)        (None, 14, 14, 512)       0         
_________________________________________________________________
conv_dw_7 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      
_________________________________________________________________
conv_dw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      
_________________________________________________________________
conv_dw_7_relu (ReLU)        (None, 14, 14, 512)       0         
_________________________________________________________________
conv_pw_7 (Conv2D)           (None, 14, 14, 512)       262144    
_________________________________________________________________
conv_pw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      
_________________________________________________________________
conv_pw_7_relu (ReLU)        (None, 14, 14, 512)       0         
_________________________________________________________________
conv_dw_8 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      
_________________________________________________________________
conv_dw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      
_________________________________________________________________
conv_dw_8_relu (ReLU)        (None, 14, 14, 512)       0         
_________________________________________________________________
conv_pw_8 (Conv2D)           (None, 14, 14, 512)       262144    
_________________________________________________________________
conv_pw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      
_________________________________________________________________
conv_pw_8_relu (ReLU)        (None, 14, 14, 512)       0         
_________________________________________________________________
conv_dw_9 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      
_________________________________________________________________
conv_dw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      
_________________________________________________________________
conv_dw_9_relu (ReLU)        (None, 14, 14, 512)       0         
_________________________________________________________________
conv_pw_9 (Conv2D)           (None, 14, 14, 512)       262144    
_________________________________________________________________
conv_pw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      
_________________________________________________________________
conv_pw_9_relu (ReLU)        (None, 14, 14, 512)       0         
_________________________________________________________________
conv_dw_10 (DepthwiseConv2D) (None, 14, 14, 512)       4608      
_________________________________________________________________
conv_dw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      
_________________________________________________________________
conv_dw_10_relu (ReLU)       (None, 14, 14, 512)       0         
_________________________________________________________________
conv_pw_10 (Conv2D)          (None, 14, 14, 512)       262144    
_________________________________________________________________
conv_pw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      
_________________________________________________________________
conv_pw_10_relu (ReLU)       (None, 14, 14, 512)       0         
_________________________________________________________________
conv_dw_11 (DepthwiseConv2D) (None, 14, 14, 512)       4608      
_________________________________________________________________
conv_dw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      
_________________________________________________________________
conv_dw_11_relu (ReLU)       (None, 14, 14, 512)       0         
_________________________________________________________________
conv_pw_11 (Conv2D)          (None, 14, 14, 512)       262144    
_________________________________________________________________
conv_pw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      
_________________________________________________________________
conv_pw_11_relu (ReLU)       (None, 14, 14, 512)       0         
_________________________________________________________________
conv_pad_12 (ZeroPadding2D)  (None, 15, 15, 512)       0         
_________________________________________________________________
conv_dw_12 (DepthwiseConv2D) (None, 7, 7, 512)         4608      
_________________________________________________________________
conv_dw_12_bn (BatchNormaliz (None, 7, 7, 512)         2048      
_________________________________________________________________
conv_dw_12_relu (ReLU)       (None, 7, 7, 512)         0         
_________________________________________________________________
conv_pw_12 (Conv2D)          (None, 7, 7, 1024)        524288    
_________________________________________________________________
conv_pw_12_bn (BatchNormaliz (None, 7, 7, 1024)        4096      
_________________________________________________________________
conv_pw_12_relu (ReLU)       (None, 7, 7, 1024)        0         
_________________________________________________________________
conv_dw_13 (DepthwiseConv2D) (None, 7, 7, 1024)        9216      
_________________________________________________________________
conv_dw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      
_________________________________________________________________
conv_dw_13_relu (ReLU)       (None, 7, 7, 1024)        0         
_________________________________________________________________
conv_pw_13 (Conv2D)          (None, 7, 7, 1024)        1048576   
_________________________________________________________________
conv_pw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      
_________________________________________________________________
conv_pw_13_relu (ReLU)       (None, 7, 7, 1024)        0         
_________________________________________________________________
global_average_pooling2d_1 ( (None, 1024)              0         
_________________________________________________________________
reshape_1 (Reshape)          (None, 1, 1, 1024)        0         
_________________________________________________________________
dropout (Dropout)            (None, 1, 1, 1024)        0         
_________________________________________________________________
conv_preds (Conv2D)          (None, 1, 1, 1000)        1025000   
_________________________________________________________________
act_softmax (Activation)     (None, 1, 1, 1000)        0         
_________________________________________________________________
reshape_2 (Reshape)          (None, 1000)              0         
=================================================================
Total params: 4,253,864
Trainable params: 4,231,976
Non-trainable params: 21,888
_________________________________________________________________
"tree_frog" 1.029875316971296e-09% confidence
"tree_frog" 0.32490063458681107% confidence
"tree_frog" 97.77554869651794% confidence

The perturbated image:

The flag:

1	picoCTF{n0w_th4t3_4_g00d_girl_2d6d5288}

Final words

The attack method used here is called fast gradient sign method. It is just like a “Hello, world!” in adversarial examples research. At this rate, machine learning might be also becoming an essential skill in CTF.

Further CTF tasks related to neural networks are below:

Task	CTF event
foolme	BCTF 2017
adamtune	DEF CON CTF Qualifier 2018
Astral Mind	SwampCTF 2018
Pilgrim	SwampCTF 2018

Let me know if there is anything else missing!

DEF CON CTF Qualifier 2018 Round-up

05-14-2018

This is too brief to be called write-up. But I’m tired …

Introduction

I’ve participated in DEF CON CTF Qualifier 2018 as a member of a certain team, ignominious 40th place. But somehow I solved 3 tasks:

ELF Crumble
babypwn1805
elastic cloud compute (memory) corruption

I write down my impressions.

ELF Crumble

This is a task to combine and execute 8 binary fragments correctly. I wrote damn brute-force solver for this, 脳が死んでいるので．

babypwn1805

A blind pwn task. I accidentally found offset -0x38 to the GOT entry of read. Then I wrote the probabilistic solver.

elastic cloud compute (memory) corruption

A VM escape task.

We were given qemu-system-x86_64 binary with vulnerable PCI device named ooo.　Notable functions are as follows:

function	description
`sub_6E61F4`	correspond to `ooo_mmio_write`
`sub_6E613C`	correspond to `ooo_mmio_read`
`sub_6E64A5`	invokes `system("cat ./flag")`

What matters is use-after-free vulnerability in:

//----- (00000000006E61F4) ----------------------------------------------------
void __fastcall sub_6E61F4(__int64 opaque, __int64 addr, __int64 value, unsigned int size)
{
  unsigned int cmd; // eax
...
  *(_QWORD *)&n[4] = value;
  cmd = ((unsigned int)addr & 0xF00000) >> 20;
  switch ( cmd )
  {
    case 1u:
          free(qword_1317940[((unsigned int)addr & 0xF0000) >> 16]);
          break;
    case 2u:
          v12 = ((unsigned int)addr & 0xF0000) >> 16;
          v8 = addr;
          memcpy((char *)qword_1317940[v12] + (signed __int16)addr, &n[4], size);
          break;
    case 0u:
          idx = ((unsigned int)addr & 0xF0000) >> 16;
          if ( idx == 15 )
          {
                for ( i = 0; i <= 14; ++i )
                  qword_1317940[i] = malloc(8LL * *(_QWORD *)&n[4]);
          }
          else
          {
                qword_1317940[idx] = malloc(8LL * *(_QWORD *)&n[4]);
          }
      break;
  }
}

With the clue of the chunk offset on 0x1317940, now we can overwrite malloc@GOT to sub_6E64A5 by fastbin attack, in particular using devmem.

I stayed up all night for this. I was tired but it was fun. I used these past write-ups as a reference when solving this task:

Thanks authors!

Final Words

Other tasks I had wanted to solve are:

flagsifier
TechSupport
smcauth

This year DEFCON’s organizer has changed from LegitBS to OOO (Order of the-Overflow). OOO seems to have the purpose of connecting academic research and CTF. I support this philosophy, but this competition was not perfect. My impressions are summarized as follows:

Pros	Cons
Meritocratic rev/pwn. Brand-new topics i.e. blockchain, neural network, reversing of Rust binary.	Many guessing tasks. Some incredible, old-fashioned tasks. In particular, sbva and ghettohackers: Throwback are quite bad.

Anyway, I’m looking forward to that next year.

Paper Gestalt Revisited: A Case Study on Computer Security Conferences

12-24-2017

HAI DOMO. This post is for 武蔵野 Advent Calendar 2017.

Introduction

In May this year, I just started my career as an apprentice security researcher at 武蔵野某所．One of my job responsibilities is to write a “good” paper that enough to be accepted to top-tier (non-crypto) security conferences like following:

IEEE S&P (Oakland)
ACM CCS
NDSS
USENIX Security

However, I am profoundly ignorant of cardinal rules of “good” security research and technical writing. ぜんぜんわからない．俺たちは雰囲気で研究をやっている．I thought I got to do something.

The joke paper entitled Paper Gestalt, distributed in CVPR’10, gave me a suggestion.

The key idea of the paper is that “good” paper might be distinguished by image recognition. だるくなってきた．時間がないので日本語で書きます．このジョーク論文では，論文を画像に変換，局所特徴量を抽出し，論文がトップカンファレンスにacceptされるかどうか判定する分類器が提案されている．仔細はPaper Gestalt - n_hidekeyの日記を参照されたい．かっこいい数式や図がある論文はそれっぽく見えてしまうよね，という話．

　そういうわけで，本稿ではPaper Gestaltを参考に，論文がセキュリティ系トップカンファレンスにacceptされるか判定する分類器を作成する．元論文ではAdaBoostを用いていたが，ここでは畳み込みニューラルネットを試す．

Dataset

　上述のカンファレンスにacceptされた論文4年分を正例，併設ワークショップにacceptされた論文同じく4年分をトップカンファレンスにrejectされた論文とみなして負例とする．負例には諸先輩方の論文が含まれていて，すみません，でもわかってくれると思うんです．

　さて一通りスクレイピングしたのち，ポスターやショートペーパーなど，4ページに満たないものを削除．キーノートやスライドも取り除く．重要なのはフルペーパーだからだ．結果，それぞれの論文数は以下の通り：

accepted	rejected
1,266	794

　正例のワードクラウド：

　負例：

　なんもわからん．

Pre-processing

　論文PDFを画像化する．

　論文PDFの各ページを横に並べ，20ページに満たない場合は白紙で埋める処理を施した．例：

　見ての通り，USENIX系の本会議に通った論文にはかっこいい表紙が付いてくる．他の論文と体裁を合わせるため削除：

　3時間ほどかけて全PDFをImageMagickで画像化，訓練用・検証用に半々で分割．

Training

　今回はベンチマークということでLeNet-5をほぼそのまま使う．いつもいつも手書き文字を認識させられるなど過酷な拷問を受けているやつ．

　フレームワークはkeras. データが少数かつclass imbalancedであることを考慮して，Building powerful image classification models using very little dataに倣い，augmentationをかけながら訓練することにした．具体的にはズームと水平方向への反転．その他各種パラメータについてはありがちな構成を雰囲気で決めている：

活性化関数	損失関数	最適化手法	Dropout	バッチサイズ	エポック	Early stopping
ReLU	クロスエントロピー	RMSProp	50%	64	100	validation accuracy

　本来ならネットワーク構成含め細かくチューニングすべきだが，手元のショボい計算機では投稿日までに計算が終わらなさそうだったため，hyperopt/hyperoptやそのkeras連携機能であるmaxpumperla/hyperasとか，そういったかっこいいテクニックは使っていない．すみません2.

Results

　Early stoppingが効いて16エポックで学習打ち止め．学習曲線：

　微妙．しかし自分が学生時代に国内研究会に投げた論文を投入したところ，

accepted	rejected	predict
7.3411%	92.6589%	rejected

とまあ正しく判定できているっぽいのでよしとしましょう．なにが正しく判定だ．俺を，馬鹿にしているのか．いま，様々なものに対して害意を抱いています．Saliency mapの可視化とかは気が向いたら．

Final Words

　ACM CCS’17のWelcome Slidesにありがたいことばが載っている：

　つまりはそういうことです．小手先の浅知恵に逃げるものはなにをやってもだめ．やるぞ〜．

DEF CON CTF Qualifier 2017 Pepperidge Farm Write-up

12-04-2017

HAI DOMO. This post is for 武蔵野 Advent Calendar 2017 and also for CTF Advent Calendar 2017.

Introduction

In May this year, I participated in DEF CON CTF Qualifier 2017 as a member of a certain 武蔵野-related team. Actually, I’m not a top-tier CTF player, but I did my best and solved 4 challenges:

crackme1
beatmeonthedl
enlightenment
Pepperidge Farm

Write-ups already exist except for Pepperidge Farm. So I decided to write about it. FYI: The binaries are available at legitbs/quals-2017.

Pepperidge Farm

Pepperidge Farm is categorized into Reverse Engineering. The problem statement is below:

Remember when the first CTF was run with a custom architecture? Pepperidge Farm remembers:
https://github.com/JonathanSalwan/VMNDH-2k12

pepperidge-farm_edb4ad2f103a9efde038346d2cd86a1e.quals.shallweplayaga.me:2012

Files
https://2017.notmalware.ru/8a45cf7171e89594cfd1d51671fef0bec3f24d9d/pepperidge_farm

It seems like a keygenning challenge on the custom virtual machine–VMNDH-2k12.

1
2
3

$ ./vmndh -file pepperidge_farm
Enter your registration code: 31337
Code does not match any known registered users

VMNDH-2k12

VMNDH-2k12 is the VM built for Nuit du Hack CTF Quals 2012 as its name suggests. The architecture is described in shell-storm | Useless emulator for fun (VMNDH-2k12). This VM parses given serialized binary and repeats fetch, decode and execution.

Writing IDA loader/processor module is a common way to analyze VM-based obfuscated binary. The modules for VMNDH-2k12 and for modified version VMNDH-2k13 already exist:

Note that when you try to solve this challenge with above-mentioned processor modules with IDA Pro 7.0, the backward-compatibility issue will occur. For example, you have to change self.regFirstSreg toself.reg_first_sreg in a module.

Also, the Binary Ninja plugin has been released after quals:

verylazyguy/binaryninja-vmndh: vmndh-2k12 Architecture plugin for Binary Ninja

But in this post, I describe a solution without both IDA Pro and Binary Ninja. Because VMNDH-2k12 is open-sourced and easy to modify.

Surface Analysis

$ ./vmndh -debug -file pepperidge_farm
[Console]#> help
<enter>              Execute next instruction
run                  Run program
bp <addr>            Set breakpoint
info bp              Display info breakpoint
info reg             Display registers
show sp              Display SP memory
show pc              Display PC memory
dis <addr>:<size>    Disassembly X bytes from ADDR
x/x <addr>:<size>    Print X bytes from ADDR
x/s <addr>           Print string addr
set reg=value        Set value in register
syscall              Execute 'syscall' instruction
help                 Display this help
quit                 Quit console debugging
[Console]#> dis
Error: Set addr between 0x8000 and 0x882a
[Console]#> dis 0x8000:82a
0x8000: jmpl 0x0720
0x8003: push r2
0x8007: .byte 0x00
0x8008: nop
...

Yes, the VM has own debugger and disassembler.

Modifying The Disassembler

However, there are pitfalls here.

Because it is a unique architecture, the destination of the control flow instructions are different from it shown on the disassembly dump. For example, take a look at src_vm/op_call.c:

static void op_call_dir16(void)
{
  uint16_t operande;
  operande = *(uint16_t *)(core.memory.vmem + core.memory.pc + 2);
  if (core.flagmode.noverbose == 0)
    fprintf(stdout, "%s0x%4.x%s > call 0x%.2x\n", RED, core.memory.pc, ENDC, operande);
  core.memory.sp -= 2; /* sub sp */
  *(uint16_t *)(core.memory.vmem + core.memory.sp) = core.memory.pc + 4;
  core.memory.pc = core.memory.pc + operande + 4;
}

Acording to this, I modified the dissassembler in src_vm/disass.c:

@@ -532,6 +537,7 @@
   uint16_t  operande;
   uint8_t   reg;
   char      flag;
+  uint16_t dst;
 
   flag = *(core.memory.vmem + addr + 1);
   switch (flag)
@@ -543,7 +549,8 @@
 
       case OP_FLAG_DIRECT16:
         operande = *(uint16_t *)(core.memory.vmem + addr + 2);
-        fprintf(stdout, "%s0x%.4x%s: call 0x%.4x\n", RED, addr, ENDC, operande);
+        dst = addr + operande + 4;
+        fprintf(stdout, "%s0x%.4x%s: call 0x%.4x\n", RED, addr, ENDC, dst);
         return (addr + 4);
 
       default:

This makes it possible to correctly display the address of the call destination in the disassembly dump. In addition, jump instructions need to be modified. In the case of jnz:

@@ -463,10 +470,11 @@
 
 static uint16_t dis_jnz(uint16_t addr)
 {
-  uint16_t operande;
+  uint16_t operande, dst;
 
   operande = *(uint16_t *)(core.memory.vmem + addr + 1);
-  fprintf(stdout, "%s0x%.4x%s: jnz 0x%.4x\n", RED, addr, ENDC, operande);
+  dst = addr + operande + 3;
+  fprintf(stdout, "%s0x%.4x%s: jnz 0x%.4x\n", RED, addr, ENDC, dst);
 
   return (addr + 3);
 }

Herewith,

0x86a1: jnz 0x0076
0x86a4: mov r0, r2
0x86a8: call 0xfdc5
0x86ac: test r0, r0
0x86af: jnz 0x0068
0x86b2: mov r0, r2
0x86b6: call 0xfddf
0x86ba: test r0, r0
0x86bd: jnz 0x005a
0x86c0: mov r0, r2
0x86c4: call 0xfe17
0x86c8: test r0, r0
0x86cb: jnz 0x004c
0x86ce: mov r0, r2
0x86d2: call 0xfe2c
0x86d6: test r0, r0
...

becomes:

0x86a1: jnz 0x871a
0x86a4: mov r0, r2
0x86a8: call 0x8471
0x86ac: test r0, r0
0x86af: jnz 0x871a
0x86b2: mov r0, r2
0x86b6: call 0x8499
0x86ba: test r0, r0
0x86bd: jnz 0x871a
0x86c0: mov r0, r2
0x86c4: call 0x84df
0x86c8: test r0, r0
0x86cb: jnz 0x871a
0x86ce: mov r0, r2
0x86d2: call 0x8502
0x86d6: test r0, r0
...

Good.

Also, after 0x8759 it looks like a data section.

0x8759: .byte 0x45 (E)
0x875a: .byte 0x6e (n)
0x875b: .byte 0x74 (t)
0x875c: .byte 0x65 (e)
0x875d: .byte 0x72 (r)
...

However, parts such as 0x87e8 are misinterpreted as codes.

1	0x87e8: inc r82

The data section was not obfuscated.

00007e0: 6c61 6700 3ce3 5a9e 9113 8c05 e42e 0a52  lag.<.Z........R
00007f0: 3dd8 7cf5 9f4b 9f06 d7a8 e9a0 a636 a649  =.|..K.......6.I
0000800: 3136 9308 eb70 ebbf df28 500d 7be8 96fa  16...p...(P.{...
0000810: e8b7 a5df c24d f0b8 3b78 6272 b748 1697  .....M..;xbr.H..
0000820: f019 2d6b 085b 9f94 4987 e624 3752 efb8  ..-k.[..I..$7R..

So I gave first aid to src_vm/disass.c:

@@ -322,8 +322,15 @@
 static uint16_t dis_inc(uint16_t addr)
 {
   uint8_t operande;
+  char OP, OP2;
 
+  OP = *(core.memory.vmem + addr);
+  OP2 = *(core.memory.vmem + addr + 1);
   operande = *(core.memory.vmem + addr + 1);
+  if(addr > 0x8759)
+    fprintf(stdout, "%s0x%.4x%s: [DEBUG] .byte 0x%.2x\n", RED, addr, ENDC, (unsigned char)OP);
+    fprintf(stdout, "%s0x%.4x%s: [DEBUG] .byte 0x%.2x\n", RED, addr + 1, ENDC, (unsigned char)OP2);
+
   fprintf(stdout, "%s0x%.4x%s: inc r%d\n", RED, addr, ENDC, operande);
 
   return (addr + 2);

Awful… who cares?

0x87e8: [DEBUG] .byte 0x0a
0x87e9: [DEBUG] .byte 0x52
0x87e8: inc r82
0x87ea: .byte 0x3d (=)

First Attempt with KLEE

This is a failure case.

As we have seen so far, VMNDH-2k12 is open-sourced. So I tried to solve the challenge with source code-based symbolic execution tool–KLEE.

I modified src_vm/syscall_write.c for assertion:

@@ -18,6 +18,8 @@
 */
 
 #include "vmndh.h"
+#include <klee/klee.h>
+#include <string.h>
 
 void syscall_write(void)
 {
@@ -32,7 +34,12 @@
 
   if (core.flagmode.noverbose == 0)
     write(1, txt, strlen(txt));
+
+  if (strstr(arg2, "Thank") != NULL)
+    klee_assert(0);
+
   core.regs.r0 = write(arg1, arg2, arg3);
+
   if (core.flagmode.noverbose == 0)
     write(1, "\n", 1);
 }

Here is a modified Makefile:

@@ -17,14 +17,15 @@
 ## Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 ##
 
-DEBUG   = no
+DEBUG   = yes
 RM      = rm -f
 INCLUDE = ./includes
 SRC_DIR = ./src_vm
+KLEE_PATH = ../klee_src/include
 NAME    = vmndh
 
 ifeq ($(DEBUG),yes)
-    CFLAGS   	= -c -ggdb3 -Wextra -Wall -D _BSD_SOURCE -I$(INCLUDE)
+    CFLAGS   	= -c -ggdb3 -Wextra -Wall -D _BSD_SOURCE -I$(INCLUDE) -I$(KLEE_PATH) -emit-llvm -g
     LDFLAGS     =
     CC 		= clang
 else

I’d left all of it to KLEE and get to bed…

$ make
$ llvm-link src_vm/*.o -o vmndh.bc
$ klee --optimize --emit-all-errors --libc=uclibc --write-smt2s --posix-runtime vmndh.bc -file pepperidge_farm -sym-stdin 65
KLEE: NOTE: Using klee-uclibc : /home/klee/klee_build/klee/Release+Debug+Asserts/lib/klee-uclibc.bca
KLEE: NOTE: Using model: /home/klee/klee_build/klee/Release+Debug+Asserts/lib/libkleeRuntimePOSIX.bca
KLEE: output directory is "/home/klee/VMNDH-2k12/klee-out-1"
KLEE: Using STP solver backend
KLEE: WARNING ONCE: function "__user_main" has inline asm
KLEE: WARNING ONCE: function "syscall_accept" has inline asm
KLEE: WARNING ONCE: function "syscall_bind" has inline asm
KLEE: WARNING ONCE: function "syscall_listen" has inline asm
KLEE: WARNING ONCE: function "syscall_recv" has inline asm
KLEE: WARNING ONCE: function "syscall_send" has inline asm
KLEE: WARNING ONCE: function "syscall_socket" has inline asm
KLEE: WARNING ONCE: calling external: syscall(16, 0, 21505, 44258832) at /home/klee/klee_src/runtime/POSIX/fd.c:1044
KLEE: WARNING ONCE: Alignment of memory from call "malloc" is not modelled. Using alignment of 8.
Enter your registration code: KLEE: WARNING ONCE: calling external: vprintf(57126208, 46345808) at /home/klee/klee_build/klee-uclibc/libc/stdio/fprintf.c:36
[SYSCALL output]: 64
KLEE: WARNING ONCE: skipping fork (memory cap exceeded)
...
KLEE: WARNING: killing 175 states (over memory cap)
Code does not match any known registered users
Code does not match any known registered users
Code does not match any known registered users
Code does not match any known registered users
Code does not match any known registered users
Code does not match any known registered users
Code does not match any known registered users
Code does not match any known registered users
Code does not match any known registered users
Code does not match any known registered users
Code does not match any known registered users
...
Code does not match any known registered users
KLEE: done: total instructions = 5117896332
KLEE: done: completed paths = 81092
KLEE: done: generated tests = 81092

… It’s not going to be easy.

I also wrote solver with angr. Which symbolizes stdin, but… let’s not talk about it.

An example of insufficient SMTLIB2 representations is:

(set-logic QF_AUFBV )
(declare-fun model_version () (Array (_ BitVec 32) (_ BitVec 8) ) )
(declare-fun stdin () (Array (_ BitVec 32) (_ BitVec 8) ) )
(declare-fun stdin-stat () (Array (_ BitVec 32) (_ BitVec 8) ) )
(assert (let ( (?B1 (select  stdin (_ bv25 32) ) ) (?B2 (select  stdin (_ bv31 32) ) ) (?B3 (select  stdin (_ bv18 32) ) ) (?B4 (select  stdin (_ bv40 32) ) ) (?B5 (select  stdin (_ bv36 32) ) ) (?B6 (select  stdin (_ bv23 32) ) ) (?B7 (select  stdin (_ bv27 32) ) ) (?B8 (select  stdin (_ bv33 32) ) ) (?B9 (select  stdin (_ bv6 32) ) ) (?B10 (select  stdin (_ bv8 32) ) ) ) (let ( (?B13 ((_ zero_extend 16)  ((_ zero_extend 8)  ?B7 ) ) ) (?B14 ((_ zero_extend 16)  ((_ zero_extend 8)  ?B8 ) ) ) (?B12 ((_ zero_extend 16)  ((_ zero_extend 8)  ?B2 ) ) ) (?B11 ((_ zero_extend 16)  ((_ zero_extend 8)  ?B1 ) ) ) ) (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (and  (=  false (=  (_ bv0 64) (bvand  (concat  (select  stdin-stat (_ bv15 32) ) (concat  (select  stdin-stat (_ bv14 32) ) (concat  (select  stdin-stat (_ bv13 32) ) (concat  (select  stdin-stat (_ bv12 32) ) (concat  (select  stdin-stat (_ bv11 32) ) (concat  (select  stdin-stat (_ bv10 32) ) (concat  (select  stdin-stat (_ bv9 32) ) (select  stdin-stat (_ bv8 32) ) ) ) ) ) ) ) ) (_ bv2147483647 64) ) ) ) (bvult  (concat  (select  stdin-stat (_ bv63 32) ) (concat  (select  stdin-stat (_ bv62 32) ) (concat  (select  stdin-stat (_ bv61 32) ) (concat  (select  stdin-stat (_ bv60 32) ) (concat  (select  stdin-stat (_ bv59 32) ) (concat  (select  stdin-stat (_ bv58 32) ) (concat  (select  stdin-stat (_ bv57 32) ) (select  stdin-stat (_ bv56 32) ) ) ) ) ) ) ) ) (_ bv65536 64) ) ) (=  (_ bv1 32) (concat  (select  model_version (_ bv3 32) ) (concat  (select  model_version (_ bv2 32) ) (concat  (select  model_version (_ bv1 32) ) (select  model_version (_ bv0 32) ) ) ) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv0 32) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv1 32) ) ) ) (=  (_ bv57 8) (select  stdin (_ bv2 32) ) ) ) (=  (_ bv70 8) (select  stdin (_ bv3 32) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv4 32) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv5 32) ) ) ) (=  false (=  (_ bv48 8) ?B9 ) ) ) (=  false (bvult  (_ bv48 32) ((_ zero_extend 16)  ((_ zero_extend 8)  ?B9 ) ) ) ) ) (=  false (=  (_ bv48 8) ?B10 ) ) ) (=  false (bvult  (_ bv48 32) ((_ zero_extend 16)  ((_ zero_extend 8)  ?B10 ) ) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv12 32) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv13 32) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv14 32) ) ) ) (=  (_ bv57 8) (select  stdin (_ bv15 32) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv16 32) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv17 32) ) ) ) (=  false (=  (_ bv48 8) ?B3 ) ) ) (=  false (bvult  (_ bv48 32) ((_ zero_extend 16)  ((_ zero_extend 8)  ?B3 ) ) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv20 32) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv21 32) ) ) ) (=  (_ bv57 8) (select  stdin (_ bv22 32) ) ) ) (=  false (=  (_ bv48 8) ?B6 ) ) ) (=  false (bvult  (_ bv48 32) ((_ zero_extend 16)  ((_ zero_extend 8)  ?B6 ) ) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv24 32) ) ) ) (=  false (=  (_ bv48 8) ?B1 ) ) ) (bvult  (_ bv48 32) ?B11 ) ) (=  false (=  (_ bv57 8) ?B1 ) ) ) (=  false (bvult  (_ bv57 32) ?B11 ) ) ) (=  (_ bv48 8) (select  stdin (_ bv26 32) ) ) ) (=  false (=  (_ bv48 8) ?B7 ) ) ) (bvult  (_ bv48 32) ?B13 ) ) (=  false (=  (_ bv57 8) ?B7 ) ) ) (=  false (bvult  (_ bv57 32) ?B13 ) ) ) (=  (_ bv48 8) (select  stdin (_ bv28 32) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv29 32) ) ) ) (=  (_ bv48 8) (select  stdin (_ bv30 32) ) ) ) (=  false (=  (_ bv48 8) ?B2 ) ) ) (bvult  (_ bv48 32) ?B12 ) ) (=  false (=  (_ bv57 8) ?B2 ) ) ) (=  false (bvult  (_ bv57 32) ?B12 ) ) ) (=  (_ bv48 8) (select  stdin (_ bv32 32) ) ) ) (=  false (=  (_ bv48 8) ?B8 ) ) ) (bvult  (_ bv48 32) ?B14 ) ) (=  false (=  (_ bv57 8) ?B8 ) ) ) (bvult  (_ bv57 32) ?B14 ) ) (=  false (=  (_ bv70 8) ?B8 ) ) ) (bvult  (_ bv70 32) ?B14 ) ) (=  false (=  (_ bv48 8) ?B5 ) ) ) (=  false (bvult  (_ bv48 32) ((_ zero_extend 16)  ((_ zero_extend 8)  ?B5 ) ) ) ) ) (=  false (=  (_ bv48 8) ?B4 ) ) ) (=  false (bvult  (_ bv48 32) ((_ zero_extend 16)  ((_ zero_extend 8)  ?B4 ) ) ) ) ) ) ) )
(check-sat)
(exit)

Solution with Z3

Since there is no choice, I read all the disassembly. After some twists and turn, I realized that:

Pepperidge Farm checks character codes against transformed 0x20 bytes values.
0x8247(x, y) returns x * 100 + y.

Now Z3 time. For example, subroutine 0x8269:

0x8269: push r1
0x826d: pop r2
0x826f: mov r7, r0
0x8273: addb r0, #00
0x8277: call 0x8247
0x827b: mov r1, r0
0x827f: mov r0, r7
0x8283: addb r0, #14
0x8287: call 0x8247
0x828b: mov r2, r0
0x828f: movl r0, #0x87ec
0x8294: call 0x8247
0x8298: xorl r1, #4936
0x829d: add r1, r2
0x82a1: xor r0, r1
0x82a5: pop r2
0x82a7: pop r1
0x82a9: ret
...
0x87ec: .byte 0x7c (|)
0x87ed: .byte 0xf5

becomes:

1	s.add(((sub_8247(0) ^ 0x4936) + sub_8247(0x14)) ^ 0x7cf5 == 0)

This is satisfiable. But not enough. We need to add rest of constraints.

#!/usr/bin/env python
#coding: utf-8
import telnetlib, sys
from z3 import *
REMOTE = len(sys.argv) >= 2 and sys.argv[1] == 'r'
if REMOTE:
    host = 'pepperidge-farm_edb4ad2f103a9efde038346d2cd86a1e.quals.shallweplayaga.me'
    port = 2012
else:
    host = '127.0.0.1'
    port = 4000
values = [BitVec('x%d' % i, 8) for i in range(0x20)]
def sub_8247(x): return ZeroExt(8, values[x]) << 8 | ZeroExt(8, values[x + 1])
s = Solver()
# sub_8269
s.add(((sub_8247(0) ^ 0x4936) + sub_8247(0x14)) ^ 0x7cf5 == 0)
# sub_82aa
s.add(sub_8247(6) * sub_8247(8) * (sub_8247(2) ^ 0xfdf) ^ 0x3dd8 == 0)
# sub_8304
s.add(((sub_8247(4) ^ 0xc7df) + sub_8247(0xe) * sub_8247(0xc)) ^ 0xeb70 == 0)
# sub_835e
s.add(((sub_8247(6) ^ 0xc5db) + 0x14aa) ^ 0x500d == 0)
# sub_838b
s.add(sub_8247(8) * sub_8247(0x1e) ^ 0x7be8 == 0)
# sub_83c7
s.add((sub_8247(0xa) + sub_8247(6) + sub_8247(0xC)) ^ 0xdf28 == 0)
# sub_841c
s.add(((sub_8247(0xC) + 0x5432) | 0x3008) ^ 0x3b78 == 0)
# sub_8449
s.add((sub_8247(0xE) + 0x1212) ^ 0x1697 == 0)
# sub_8471
s.add((sub_8247(0x10) ^ 0x8703) ^  0x3136 == 0)
# sub_8499
s.add(((sub_8247(0x12) + 0x4004) + (sub_8247(0x14) ^ 0xA52)) ^  0x6272 == 0)
# sub_84df
s.add()
# sub_8502
s.add((sub_8247(0x16) + sub_8247(0x10)) ^ 0x9308 == 0)
# sub_853e
s.add(sub_8247(0x18) ^ 0x085b == 0)
# sub_8561
s.add(((sub_8247(0x1a) ^ 0x863c) + 0x1234) ^ 0x9113 == 0)
# sub_858e
s.add((sub_8247(0x1c) + sub_8247(8) + sub_8247(0x12)) ^ 0xf0b8 == 0)
# sub_85e3
s.add(((sub_8247(0x1e) & 0x0f00) + sub_8247(0)) ^ 0x9f94 == 0)
print s.check()
print s.model()
input = ''
for i in range(0x20):
    try:
        input += '%02x' % s.model()[values[i]].as_long()
    except:
        input += '??'
input = input.upper()
print input
t = telnetlib.Telnet(host, port)
t.write(input + '\n')
t.interact()

Even with halfway constraints, the process proceeds. So inscount with Pin or other dynamic binary instrumentation tools might be helpful.

Finally I got:

Enter your registration code: 90940C6017E5FEB82083F932E73E0485B635796DA353DCD3085BF8E356C87FF8
Thank you for your patronage!
Your username is: pepperidge
This is the flag: th0s3_wh0_l0ok_0nly_to_th3_p4stur3_ar3_cert4in_t0_miss_th3_futur3

The conclusive SMTLIB2 representation is:

; benchmark
(set-info :status unknown)
(set-logic QF_AUFBV)
(declare-fun x1 () (_ BitVec 8))
(declare-fun x0 () (_ BitVec 8))
(declare-fun x31 () (_ BitVec 8))
(declare-fun x30 () (_ BitVec 8))
(declare-fun x19 () (_ BitVec 8))
(declare-fun x18 () (_ BitVec 8))
(declare-fun x9 () (_ BitVec 8))
(declare-fun x8 () (_ BitVec 8))
(declare-fun x29 () (_ BitVec 8))
(declare-fun x28 () (_ BitVec 8))
(declare-fun x27 () (_ BitVec 8))
(declare-fun x26 () (_ BitVec 8))
(declare-fun x25 () (_ BitVec 8))
(declare-fun x24 () (_ BitVec 8))
(declare-fun x17 () (_ BitVec 8))
(declare-fun x16 () (_ BitVec 8))
(declare-fun x23 () (_ BitVec 8))
(declare-fun x22 () (_ BitVec 8))
(declare-fun x21 () (_ BitVec 8))
(declare-fun x20 () (_ BitVec 8))
(declare-fun x15 () (_ BitVec 8))
(declare-fun x14 () (_ BitVec 8))
(declare-fun x13 () (_ BitVec 8))
(declare-fun x12 () (_ BitVec 8))
(declare-fun x7 () (_ BitVec 8))
(declare-fun x6 () (_ BitVec 8))
(declare-fun x11 () (_ BitVec 8))
(declare-fun x10 () (_ BitVec 8))
(declare-fun x5 () (_ BitVec 8))
(declare-fun x4 () (_ BitVec 8))
(declare-fun x3 () (_ BitVec 8))
(declare-fun x2 () (_ BitVec 8))
(assert
 (let ((?x206 (bvand (bvor (bvshl ((_ zero_extend 8) x30) (_ bv8 16)) ((_ zero_extend 8) x31)) (_ bv3840 16))))
(let ((?x209 (bvxor (bvadd ?x206 (bvor (bvshl ((_ zero_extend 8) x0) (_ bv8 16)) ((_ zero_extend 8) x1))) (_ bv40852 16))))
(let (($x210 (= ?x209 (_ bv0 16))))
(let ((?x82 (bvor (bvshl ((_ zero_extend 8) x8) (_ bv8 16)) ((_ zero_extend 8) x9))))
(let ((?x200 (bvadd (bvor (bvshl ((_ zero_extend 8) x28) (_ bv8 16)) ((_ zero_extend 8) x29)) ?x82)))
(let ((?x201 (bvadd ?x200 (bvor (bvshl ((_ zero_extend 8) x18) (_ bv8 16)) ((_ zero_extend 8) x19)))))
(let (($x204 (= (bvxor ?x201 (_ bv61624 16)) (_ bv0 16))))
(let ((?x190 (bvxor (bvor (bvshl ((_ zero_extend 8) x26) (_ bv8 16)) ((_ zero_extend 8) x27)) (_ bv34364 16))))
(let (($x195 (= (bvxor (bvadd ?x190 (_ bv4660 16)) (_ bv37139 16)) (_ bv0 16))))
(let ((?x183 (bvxor (bvor (bvshl ((_ zero_extend 8) x24) (_ bv8 16)) ((_ zero_extend 8) x25)) (_ bv2139 16))))
(let (($x184 (= ?x183 (_ bv0 16))))
(let ((?x174 (bvadd (bvor (bvshl ((_ zero_extend 8) x22) (_ bv8 16)) ((_ zero_extend 8) x23)) (bvor (bvshl ((_ zero_extend 8) x16) (_ bv8 16)) ((_ zero_extend 8) x17)))))
(let (($x177 (= (bvxor ?x174 (_ bv37640 16)) (_ bv0 16))))
(let ((?x165 (bvxor (bvor (bvshl ((_ zero_extend 8) x20) (_ bv8 16)) ((_ zero_extend 8) x21)) (_ bv2642 16))))
(let ((?x163 (bvadd (bvor (bvshl ((_ zero_extend 8) x18) (_ bv8 16)) ((_ zero_extend 8) x19)) (_ bv16388 16))))
(let (($x169 (= (bvxor (bvadd ?x163 ?x165) (_ bv25202 16)) (_ bv0 16))))
(let ((?x154 (bvxor (bvor (bvshl ((_ zero_extend 8) x16) (_ bv8 16)) ((_ zero_extend 8) x17)) (_ bv34563 16))))
(let (($x157 (= (bvxor ?x154 (_ bv12598 16)) (_ bv0 16))))
(let ((?x145 (bvadd (bvor (bvshl ((_ zero_extend 8) x14) (_ bv8 16)) ((_ zero_extend 8) x15)) (_ bv4626 16))))
(let (($x148 (= (bvxor ?x145 (_ bv5783 16)) (_ bv0 16))))
(let ((?x107 (bvor (bvshl ((_ zero_extend 8) x12) (_ bv8 16)) ((_ zero_extend 8) x13))))
(let (($x143 (= (bvxor (bvor (bvadd ?x107 (_ bv21554 16)) (_ bv12296 16)) (_ bv15224 16)) (_ bv0 16))))
(let ((?x78 (bvor (bvshl ((_ zero_extend 8) x6) (_ bv8 16)) ((_ zero_extend 8) x7))))
(let ((?x132 (bvadd (bvor (bvshl ((_ zero_extend 8) x10) (_ bv8 16)) ((_ zero_extend 8) x11)) ?x78)))
(let (($x136 (= (bvxor (bvadd ?x132 ?x107) (_ bv57128 16)) (_ bv0 16))))
(let ((?x124 (bvmul ?x82 (bvor (bvshl ((_ zero_extend 8) x30) (_ bv8 16)) ((_ zero_extend 8) x31)))))
(let (($x127 (= (bvxor ?x124 (_ bv31720 16)) (_ bv0 16))))
(let (($x119 (= (bvxor (bvadd (bvxor ?x78 (_ bv50651 16)) (_ bv5290 16)) (_ bv20493 16)) (_ bv0 16))))
(let ((?x108 (bvmul (bvor (bvshl ((_ zero_extend 8) x14) (_ bv8 16)) ((_ zero_extend 8) x15)) ?x107)))
(let ((?x109 (bvadd (bvxor (bvor (bvshl ((_ zero_extend 8) x4) (_ bv8 16)) ((_ zero_extend 8) x5)) (_ bv51167 16)) ?x108)))
(let (($x112 (= (bvxor ?x109 (_ bv60272 16)) (_ bv0 16))))
(let ((?x90 (bvmul (bvmul ?x78 ?x82) (bvxor (bvor (bvshl ((_ zero_extend 8) x2) (_ bv8 16)) ((_ zero_extend 8) x3)) (_ bv4063 16)))))
(let (($x93 (= (bvxor ?x90 (_ bv15832 16)) (_ bv0 16))))
(let ((?x50 (bvadd (bvxor (bvor (bvshl ((_ zero_extend 8) x0) (_ bv8 16)) ((_ zero_extend 8) x1)) (_ bv18742 16)) (bvor (bvshl ((_ zero_extend 8) x20) (_ bv8 16)) ((_ zero_extend 8) x21)))))
(let (($x54 (= (bvxor ?x50 (_ bv31989 16)) (_ bv0 16))))
(and $x54 $x93 $x112 $x119 $x127 $x136 $x143 $x148 $x157 $x169 $x177 $x184 $x195 $x204 $x210)))))))))))))))))))))))))))))))))))))
(check-sat)

Final Words

I enjoyed this challenge. It seems to be easy or medium-easy difficulty rating. If only I could have solved more difficult Reverse Engineering challenges during quals–liberty, godzilla, and so on.

Recently I read T. Blazytko et al. USENIX Security’17. The paper says the system named Syntia automatically deobfuscate binaries with program synthesis. Program synthesis is a method to synthesize some pieces of program from given I/O samples and possible operators–like this:

#!/usr/bin/env python
#coding: utf-8
from z3 import *
x, y, h = BitVecs(['x', 'y', 'h'], 8)
# I/O samples
t1 = And(x == 1, y == 2)
t2 = And(x == 2, y == 4)
t3 = And(x == 3, y == 6)
phi_1 = Or(t1, t2, t3)
# h is subject to synthesize
phi_2 = y == x << h
s = Solver()
s.add(ForAll([x, y], Implies(phi_1, phi_2)))
print s.check() # sat
print s.model() # [h = 2]

This is just a simple example. In practical, program slicing and path pruning will be needed. Both symbolic execution and program synthesis depend on SMT solver, but according to the paper, program synthesis is more suitable for deobfuscation tasks… really? I’ll investigate further.

How I Found the Honeypot: Dark Web OSINT and Image Processing

12-02-2017

This post is for Honeypot Advent Calendar 2017.

Introduction

In May this year, Trend Micro researchers have announced interesting research results in the article titled Red on Red: The Attack Landscape of the Dark Web - TrendLabs Security Intelligence Blog. They had deployed a honeypot on the dark web and monitored attack activities. They’ve done great work, indeed.

Well then, with the aid of the screenshot in that article, probably I found the honeypot. May I introduce the how and why?

Dark Web OSINT

In March this year, when they would have created an presentation slide, I have been running the crawler for the dark web. My purpose was to create a pictorial book of Tor hidden services below:

The crawler is simple, just like saying “Hello, world!” to PhantomJS. It has only capability of taking a screenshot.

service_args = [ 
            # Do not insert blank to each of args.
            '--proxy=127.0.0.1:9050',
            '--proxy-type=socks5'
        ]
dcap = {
        'phantomjs.page.settings.userAgent': 'Mozilla/5.0 (Windows NT 6.1; rv:31.0) Gecko/20100101 Firefox/31.0'
}
def get_title_with_screenshot(url):
    driver = webdriver.PhantomJS(service_args = service_args, desired_capabilities = dcap)
    driver.set_window_size(1024, 512)
    driver.get('http://' + url + '.onion') # 'http://' is required.
    driver.save_screenshot(url + '.png')
    title = driver.title
    driver.close()
    return title

I have discovered 40,208 onion domains and confirmed 1,797 domains were active.

Image Processing

Thanks to collecting screenshots by chance, I was able to find a site similar to the screenshot in their article–with histgram calculation:

list = glob.glob('./images/*.png')
def cbir():
    target_im = cv2.imread(sys.argv[1])
    target_hist = cv2.calcHist([target_im], [0], None, [256], [0, 256])
    for i in list:
        comparing_im = cv2.imread(i)
        comparing_hist = cv2.calcHist([comparing_im], [0], None, [256], [0, 256])
        diff = cv2.compareHist(target_hist, comparing_hist, 0)
        if diff > float(sys.argv[2]): # Threshold
            print i, diff

Yet domain names are not posted in their articles, I believe this is it.

$ wget http://blog.trendmicro.com/trendlabs-security-intelligence/files/2017/05/red-on-red-1.jpg
$ cbir.py red-on-red-1.jpg
$ python cbir.py red-on-red-1.jpg 0.9
./images/s5**********jlp2.png 0.979927357262

The screenshot:

There are many clone sites in the dark web–for backup, or even for spying? s5**********jlp2.onion might be cloned and have maintained by others. Even in that case, the site is likely to be a honeypot.

Interestingly, I also found some posts like to induce to the site at r/onions. I believe these are done by researcher.

Last Words

We got a glimpse of deep in abyss. This is just an accidental case study. Needless to say, no insult intended.

My crawler and image processing scripts are available at ntddk/onionstack. Crawling of the dark web is accompanied by risk. After all, with ethical considerations, I’ve deleted screenshots I’d captured except for the honeypot.

If you interested in the dark web OSINT, Dark Web | Automating OSINT Blog will be a good starting point.

Anyway, keep safety.

Backward Program Slicing with IDAPython and Triton

11-23-2017

Introduction

As you know, IDAPython is quite useful. And Triton concolic execution engine has python binding. Then… why not integrate them? I tried to stand on the shoulders of giants.

Backward Program Slicing

Roughly speaking, program slicing is a method to extract subset of program which is relevant to given statement. Here is an excerpt from M. Weiser. ICSE’81:

Starting from a subset of a program’s behavior, slicing reduces that program to a minimal form which still produces that behavior. The reduced program, called a “slice”, is an independent program guaranteed to faithfully represent the original program within the domain of the specified subset of behavior.

Kudos to Jonathan Salwan, we can easily apply backward program slicing to binary analysis process with minor modification of backward_slicing.py and proving_opaque_predicates.py. I wrote a simple, tiny glue between Triton and IDA Pro:

#!/usr/bin/env python
#coding: utf-8
from idc import *
from idaapi import *
from triton import *
ctx = TritonContext()
regs = {
        'eax': REG.X86.EAX,
        'ebx': REG.X86.EBX,
        'ecx': REG.X86.ECX,
        'edx': REG.X86.EDX,
        }
def symbolization_init():
    ctx.convertRegisterToSymbolicVariable(ctx.registers.eax)
    ctx.convertRegisterToSymbolicVariable(ctx.registers.ebx)
    ctx.convertRegisterToSymbolicVariable(ctx.registers.ecx)
    ctx.convertRegisterToSymbolicVariable(ctx.registers.edx)
    return
def opaque_predicate_detection(pc):
    ctx.setArchitecture(ARCH.X86)
    ctx.setAstRepresentationMode(AST_REPRESENTATION.PYTHON)
    symbolization_init()
    ast_ctx = ctx.getAstContext()
    while True:
        instruction = Instruction()
        instruction.setAddress(pc)
        opcode = GetManyBytes(pc, ItemSize(pc)) # Disassemble with IDA Pro
        instruction.setOpcode(opcode)
        ctx.processing(instruction) # Emulate instruction without Pintool
        for se in instruction.getSymbolicExpressions():
            se.setComment(str(instruction))
        if str(instruction.getDisassembly()).split()[0] == 'cmp': # if instruction.isCompare():
            reg = str(instruction.getOperands()[0]).split(':')[0] # Get first operand
            if reg in regs:
                Expr = ctx.getSymbolicRegisters()[regs[reg]]
                slicing = ctx.sliceExpressions(Expr)
                for k, v in sorted(slicing.items()):
                    print v.getComment()
                print instruction
        elif instruction.isBranch():
            print instruction
            if instruction.isConditionTaken():
                print 'branch will taken'
            else:
                print 'branch will not taken'
            break
        pc = NextHead(pc)
    print '-' * 50
def main():
    ea = ScreenEA() # Get address corresponding to cursor
    opaque_predicate_detection(ea) 
if __name__ == '__main__':
    view = main()

Showcase

The snippet extracts subset of program which is relevant to branch condition. We can run this from File -> Script file in IDA Pro menu.

Conditional Branch

.text:004074DD     loc_4074DD:
.text:004074DD 040 mov     dword ptr [edi+18h], 0Ah
.text:004074E4 040 mov     dword ptr [edi+14h], 0
.text:004074EB 040 mov     word ptr [edi+8], 8235h
.text:004074F1 040 mov     dword ptr [edi+0Ch], 3E8h
.text:004074F8 040 mov     dword ptr [edi+10h], 0BB8h
.text:004074FF 040 sub     esp, 4
.text:00407502 044 mov     [esp+40h+Memory], 0B8h ; Size        ; cursor
.text:00407509 044 call    ??2@YAPAXI@Z    ; operator new(uint)
.text:0040750E 044 add     esp, 4
.text:00407511 040 mov     [esi+18h], eax
.text:00407514 040 mov     eax, dword_591EE4
.text:00407519 040 mov     ecx, dword_591EE0
.text:0040751F 040 imul    eax, eax
.text:00407522 040 imul    edx, eax, 7
.text:00407525 040 dec     edx
.text:00407526 040 mov     eax, ecx
.text:00407528 040 imul    eax, eax
.text:0040752B 040 cmp     edx, eax
.text:0040752D 040 jz      short loc_407538

becomes:

0x407514: mov eax, dword ptr [0x591ee4]
0x40751f: imul eax, eax
0x407522: imul edx, eax, 7
0x407525: dec edx
0x40752b: cmp edx, eax
0x40752d: je 0x407538
branch will not taken

Unconditional Branch

.text:004078C3     loc_4078C3:
.text:004078C3 000 mov     esp, edi           ; cursor
.text:004078C5 000 mov     eax, dword_591EAC
.text:004078CA 000 lea     ecx, [eax+4]
.text:004078CD 000 mov     edx, eax
.text:004078CF 000 sar     edx, cl
.text:004078D1 000 lea     eax, [eax+edx*2]
.text:004078D4 000 mov     dword_591EAC, eax
.text:004078D9 000 jmp     short loc_4

becomes:

1 2	0x4078d9: jmp 0x40789a branch will taken

Looks nice.

Last Words

Triton’s emulation iteration is compatible to IDAPython manner. Therefore, The combination of IDA Pro and Triton is pretty good.

Cheers,

A Supplemental Guide to Migrate PySide Code to PyQt5 Within Your IDAPython Script

11-08-2017

Introduction

IDAPython is a powerful feature of IDA Pro, and there are many open-sourced IDAPython projects. However, we cannot use every GUI-based IDAPython script due to some Qt-related breaking changes between IDA Pro 6.8 and 6.9 or later. The main problem is about migrating no longer supported PySide code to PyQt5.

Recently I ported PySide code within idasec–one of the most sophisticated deobfuscation frameworks, which tackles opaque predicates and call stack tampering in terms of infeasibility questions, by utilizing Backward-Bounded Dynamic Symbolic Execution proposed in the remarkably well written paper S. Bardin et al. IEEE S&P’17–to PyQt5.

That’s why I decided to write this blog post for a note to self and for someone trying to do similar thing.

There are 2 guidances to migrate PySide code to PyQt5:

Please read them before. I only give supplemental information in addition to predecessors.

How to Migrate

Now let’s get started.

Change `QtGui` methods to `QtWidgets`

Most methods in QtGui migrated to QtWidgets. Therefore,

1	from PySide import QtGui, QtCore

becomes:

1	from PyQt5 import QtCore, QtGui, QtWidgets

As an example, QTextEdit described in Hex Blog. In additions, the methods to be rewritten are as follows:

QtWidgets.QLayout
QtWidgets.QVBoxLayout
QtWidgets.QHBoxLayout
QtWidgets.QWidget
QtWidgets.QTableWidget
QtWidgets.QListWidget
QtWidgets.QTabWidget
QtWidgets.QDockWidget
QtWidgets.QTreeWidget
QtWidgets.QTreeWidgetItem
QtWidgets.QPushButton
QtWidgets.QRadioButton
QtWidgets.QToolButton
QtWidgets.QButtonGroup
QtWidgets.QGroupBox
QtWidgets.QSpinBox
QtWidgets.QCheckBox
QtWidgets.QComboBox
QtWidgets.QTextEdit
QtWidgets.QLineEdit
QtWidgets.QApplication
QtWidgets.QLabel
QtWidgets.QSizePolicy
QtWidgets.QMenu
QtWidgets.QFrame
QtWidgets.QProgressBar
QtWidgets.QStyle
QtWidgets.QSpacerItem
QtWidgets.QScrollArea
QtWidgets.QSplitter
There might be more…

My experience says that other than the following 3 methods may be rewritten:

QtGui.QPixmap
QtGui.QIcon
QtGui.QFont

idacute may overwrite all of QtGui methods, so I think there still needs to be manual works.

Overwrite `_fromUtf8`

We also need to overwrite _fromUtf8.

try:
    _fromUtf8 = QtCore.QString.fromUtf8
except AttributeError:
　　　　_fromUtf8 = lambda s: s

Others

These issues are described by predecessors:

Handling SIGNAL
Change FormToPySideWidget to FormToPyQtWidget
Change setResizeMode to setSectionResizeMode

Conclusion

This time, I was able to run idasec on IDA Pro 7.0 with some bug fixes and dirty patches – like this cool video:

If you are an IDA Pro 7.0 user, note that other backward-compatibility issue described in IDA: IDAPython backward-compatibility with 6.95 APIs will occur.

Enjoy!

HAI DOMO VIRTUAL YOUTUBER KIZUNA AI DESU. I’m still working on my English.

死にゆくアンチウイルスへの祈り

09-10-2017

　Security meets Machine Learningという勉強会にて，上記のタイトルで発表した．資料はこちら：

　謎の力が働いて会社からの発表になっておりますが，機械学習の研究をしているわけではありません．既存研究の再現実装を試みているとこれ中国語の部屋じゃんという気持ちになる．
　ともあれ，これまで各種資料はただSpeakerDeckに載せるだけだったのを今後はブログから一元的に参照できるようにします．

z3pyリンク集

08-24-2017

　なにもかも忘れかけている．

はじめに

　z3pyはSMTソルバZ3のPythonバインディング．たとえば

...
    if(x * 2 + 3 * y == 4
        && x * 3 + y == -1){
        puts("congrats!");
    }

のようなプログラムに対して

from z3 import *
x = Int('x')
y = Int('y')
s = Solver()
s.add(x * 2 + 3 * y == 4)
s.add(x * 3 + y == -1)
print(s.check())
print(s.model())

としてやれば期待される入力値[x = -1, y = 2]を算出できる．これだけではだからなに？という感じだが．シンボリック実行は各命令とメモリの状態からこの割当を自動生成するものだと思えばよい．これがangrやらTritonやらmiasmやらmanticoreやら，その他いまどきのバイナリ解析ツール群の基盤となっているというわけです．

参考になったサイト

　で，メモっとかないと忘れるので．SAT/SMTソルバのしくみやシンボリック実行全般まで広げると膨大になるということで，z3pyに限っています．前者については高速SATソルバーの原理 - 基盤(S)離散構造処理系プロジェクト[PDF]やSAT/SMTソルバの仕組み - SlideShareを読むとよいでしょう．いまどきのSAT/SMTソルバにはVSIDSやらLubyリスタートやらいろいろな工夫が盛り込まれているが，とりあえずユーザとしてはDPLLとCDCLさえ抑えておけば問題ないはず．後者，シンボリック実行全般に関するおすすめの文献は秘密．Pythonバインディング以外の記事，angrやTritonなどサードパーティのツール群に関する記事も省いてある．

ウェブサイト	概要
CTF/Toolkit/z3py - 電気通信大学MMA	日本語で書かれた入門資料．これには書かれてないが`BitVecSort`も便利．
Theorem prover, symbolic execution and practical reverse-engineering	z3pyのチュートリアル．まずはこの通りに手を動かす．
Breaking Kryptonite’s Obfuscation: A Static Analysis Approach Relying on Symbolic Execution	シンボリック実行の最小構成の実装．
0vercl0k/z3-playground	上2つの資料で参照されているソースコード群．
Solving a simple crackme using Z3 – Aneesh Dogra’s Blog	簡単なcrackmeのwriteup.
Playing with Z3, hacking the serial check. – rosako’s blog	簡単なcrackmeのwriteup.
Using SAT and SMT to defeat simple hashing algorithms - LSE Blog	オレオレハッシュ関数の解析．
Reversing the petya ransomware with constraint solvers	ランサムウェアPetyaのSalsa実装の不備を突くdecryptor.
thomasjball/PyExZ3	シンボリック実行の実装例．z3公式からリンクが貼られている．`symbolic/`以下，`z3_wrap.py`, `loader.py`が参考になる．
Quick introduction into SAT/SMT solvers and symbolic execution (DRAFT) [PDF]	いろいろリンク貼ったけどこれだけ読めばいい．SMTソルバの紹介．数独の解き方．マインスイーパの自動化．デコンパイラ．難読化解除．シンボリック実行．ハッシュ関数の解析．全部載っている．Dennis Yurichev氏，Reverse Engineering for Beginnersも書いていて，慈善事業家か？

　はてなブックマークを使っていればよかったのではという気がしてきた．

おわりに

　z3py便利最高ですね．みなさんは便利最高ですか？
　これで次のようなCTFの問題を解くことができます：

問題	大会
Unbreakable Enterprise Product Activation	Google CTF 2016
Ropsynth	SECCON 2016 Online CTF
baby-re	DEF CON CTF Qualifier 2016
amandhj	DEF CON CTF Qualifier 2016
Pepperidge Farm	DEF CON CTF Qualifier 2017

　まあ明らかにangr使ったほうが楽．とりあえずいくつかやってみただけで，ほかにもたくさんあると思う．
　Bareflank Hypervisorを読むのはいつになるかわかりません．

Introduction

Solution

Final words

Introduction

ELF Crumble

babypwn1805

elastic cloud compute (memory) corruption

Final Words

Introduction

Dataset

Pre-processing

Training

Results

Final Words

Introduction

Pepperidge Farm

VMNDH-2k12

Surface Analysis

Modifying The Disassembler

First Attempt with KLEE

Solution with Z3

Final Words

Introduction

Dark Web OSINT

Image Processing

Last Words

Introduction

Backward Program Slicing

Showcase

Conditional Branch

Unconditional Branch

Last Words

Introduction

Related Work

How to Migrate

Change QtGui methods to QtWidgets

Overwrite _fromUtf8

Others

Conclusion

はじめに

参考になったサイト

おわりに

Change `QtGui` methods to `QtWidgets`

Overwrite `_fromUtf8`