saketupadhyay
commited on
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,128 @@
|
|
1 |
---
|
2 |
license: mit
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: mit
|
3 |
+
task_categories:
|
4 |
+
- tabular-classification
|
5 |
+
tags:
|
6 |
+
- software
|
7 |
+
size_categories:
|
8 |
+
- 10K<n<100K
|
9 |
---
|
10 |
+
|
11 |
+
# FuzzDistill Dataset Explanation
|
12 |
+
|
13 |
+
The final Basic Block and Function features are extracted in Semicolon-Separated Values (SSV) format.
|
14 |
+
|
15 |
+
To load in pandas, use -
|
16 |
+
```python
|
17 |
+
import pandas as pd
|
18 |
+
data = pd.read_csv("FNFeatures.csv", sep=";")
|
19 |
+
```
|
20 |
+
Assuming `FNFeatures.csv` is the target feature file.
|
21 |
+
|
22 |
+
## Function Features
|
23 |
+
|
24 |
+
The generated dataset is list of functions with various characteristics and a label indicating whether each function is
|
25 |
+
vulnerable or not. The data is structured into 14 columns, which are described below:
|
26 |
+
|
27 |
+
1. **Function ID**: A unique identifier for each function.
|
28 |
+
2. **Function Name**: The name of the function.
|
29 |
+
3. **Instructions**: The number of instructions (e.g.,Intermediate Instructions) in the function.
|
30 |
+
4. **BBs** (Basic Blocks): The number of basic blocks in the function. A basic block is a sequence of instructions that
|
31 |
+
are executed together without any control flow changes.
|
32 |
+
5. **In-degree**: The number of incoming edges to the function in the call graph, indicating how many other functions
|
33 |
+
call this one.
|
34 |
+
6. **Out-degree**: The number of outgoing edges from the function in the call graph, indicating how many other functions
|
35 |
+
are called by this one.
|
36 |
+
7. **Num Loops**: The number of loops (e.g., for, while, do-while) present in the function.
|
37 |
+
8. **Static Allocations**: The number of static memory allocations made by the function.
|
38 |
+
9. **Dynamic Allocations**: The number of dynamic memory allocations made by the function (e.g., using `malloc`,
|
39 |
+
`realloc`).
|
40 |
+
10. **MemOps** (Memory Operations): The number of memory-related operations performed by the function (e.g., reads,
|
41 |
+
writes).
|
42 |
+
11. **CondBranches** (Conditional Branches): The number of conditional branches (e.g., if-else statements) in the
|
43 |
+
function.
|
44 |
+
12. **UnCondBranches** (Unconditional Branches): The number of unconditional branches (e.g., jumps, returns) in the
|
45 |
+
function.
|
46 |
+
13. **DirectCalls**: The number of direct function calls made by the function.
|
47 |
+
14. **InDirectCalls** (Indirect Calls): The number of indirect function calls made by the function (e.g., through a
|
48 |
+
pointer or a table).
|
49 |
+
15. **VULNERABLE**: A binary label indicating whether the function is vulnerable (1) or not (0).
|
50 |
+
|
51 |
+
## Basic Block Features
|
52 |
+
|
53 |
+
Generated Basic Block dataset a collection of basic blocks (BBs) from functions with various characteristics and a label
|
54 |
+
indicating whether each block is vulnerable or not. The data is structured into 13 columns, which are described below:
|
55 |
+
|
56 |
+
1. **Block ID**: A unique identifier for each basic block.
|
57 |
+
2. **Block Name**: Name of the block with following structure - `BB_<block #>_<demangled parent function>`
|
58 |
+
3. **Instructions**: The number of instructions (e.g., assembly code operations) in the basic block.
|
59 |
+
4. **In-degree**: The number of incoming edges to the basic block in the control flow graph, indicating how many other
|
60 |
+
blocks lead to this one.
|
61 |
+
5. **Out-degree**: The number of outgoing edges from the basic block in the control flow graph, indicating how many
|
62 |
+
other blocks are reachable from this one.
|
63 |
+
6. **Static Allocations**: The number of static memory allocations made by the basic block.
|
64 |
+
7. **Dynamic Allocations**: The number of dynamic memory allocations made by the basic block (e.g., using `new`,
|
65 |
+
`malloc`).
|
66 |
+
8. **MemOps** (Memory Operations): The number of memory-related operations performed by the basic block (e.g., reads,
|
67 |
+
writes).
|
68 |
+
9. **CondBranches** (Conditional Branches): The number of conditional branches (e.g., if-else statements) in the basic
|
69 |
+
block.
|
70 |
+
10. **UnCondBranches** (Unconditional Branches): The number of unconditional branches (e.g., jumps, returns) in the
|
71 |
+
basic block.
|
72 |
+
11. **DirectCalls**: The number of direct function calls made by the basic block.
|
73 |
+
12. **InDirectCalls** (Indirect Calls): The number of indirect function calls made by the basic block (e.g., through a
|
74 |
+
pointer or a table).
|
75 |
+
13. **VULNERABLE**: A binary label indicating whether the basic block is vulnerable (1) or not (0).
|
76 |
+
|
77 |
+
---
|
78 |
+
|
79 |
+
### A Note on Branches in Basic Blocks
|
80 |
+
|
81 |
+
Conditional branches (CondBranches) and unconditional branches (UnCondBranches) primarily serve as sanity checks and do
|
82 |
+
not significantly impact the categorization of Basic Blocks (it might actually harm the accuracy). Let’s analyze the
|
83 |
+
possible values of \( N \) (number of conditional branches) and \( M \) (number of unconditional branches).
|
84 |
+
|
85 |
+
A basic block can contain at most one conditional branch. A conditional branch is typically used to terminate the block
|
86 |
+
and transfer control to another location within the code. If there were multiple conditional branches, they would need
|
87 |
+
to be combined into a single decision point using logical operators, which would not increase the count of separate
|
88 |
+
conditional branches.
|
89 |
+
|
90 |
+
$$
|
91 |
+
\therefore N \in \{0, 1\}
|
92 |
+
$$
|
93 |
+
|
94 |
+
where \( N \) is either \( 0 \) (no conditional branch) or \( 1 \) (one conditional branch).
|
95 |
+
|
96 |
+
Similarly, a basic block can have at most one unconditional branch. An unconditional branch is typically used to exit
|
97 |
+
the block and jump to another location in the code. If there were multiple unconditional branches, they would be
|
98 |
+
redundant, as only one of them would be executed.
|
99 |
+
|
100 |
+
$$
|
101 |
+
\therefore M \in \{0, 1\}
|
102 |
+
$$
|
103 |
+
|
104 |
+
where \( M \) is either \( 0 \) (no unconditional branch) or \( 1 \) (one unconditional branch).
|
105 |
+
|
106 |
+
If a basic block contains a conditional branch (\( N = 1 \)), it is not possible to have an unconditional branch (\( M =
|
107 |
+
0 \)), as the control flow would be determined solely by the conditional branch. Conversely, if a basic block includes
|
108 |
+
an unconditional branch (\( M = 1 \)), it is not feasible to have a conditional branch (\( N = 0 \)), as the
|
109 |
+
unconditional branch would override any conditional decision.
|
110 |
+
|
111 |
+
Logically -
|
112 |
+
|
113 |
+
$$
|
114 |
+
N \times M = 0
|
115 |
+
$$
|
116 |
+
|
117 |
+
$$
|
118 |
+
(N = 1) \Rightarrow (M = 0)
|
119 |
+
$$
|
120 |
+
|
121 |
+
$$
|
122 |
+
(M = 1) \Rightarrow (N = 0)
|
123 |
+
$$
|
124 |
+
|
125 |
+
That means that only one of \( N \) or \( M \) can have the value of \( 1 \) at any given time.
|
126 |
+
If \( N \) is set to \( 1 \), \( M \) must be set to \( 0 \), and vice versa.
|
127 |
+
|
128 |
+
We can use this relationship to check the functionality of our BB compiler pass and sanity of our training dataset.
|