> For the complete documentation index, see [llms.txt](https://alham-rizvi.gitbook.io/alhamrizvi/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://alham-rizvi.gitbook.io/alhamrizvi/picoctf-writeups/picoctf/forensics/investigation_encoded-1.md).

# Investigation Encoded 1

## Challenge Overview

We are given a binary called `mystery` and an `output` file. The goal is to recover the original flag that was encoded into `output`.

***

## Step 1: Initial Recon

First, check what the `output` file actually is:

```bash
file output
```

Result: `Non-ISO extended-ASCII text, with no line terminators` — not an image or plaintext flag.

Inspect the raw bytes:

```bash
xxd -g 1 output
```

```
00000000: 8e 8e ba 3b b8 ea 23 a8  a8 ba 3a 8e ee 38 ee a3  ...;..#...:..8..
00000010: a8 eb 8a a3 ab 80                                  ......
```

The output is short and clearly encoded — not human-readable.

***

## Step 2: Running the Binary

```bash
./mystery
```

Output: `./flag.txt not found`

Create a fake flag (no newline, no underscore — we'll learn why shortly):

```bash
echo -n picoCTF{fakeflag} > flag.txt
./mystery
```

Output: `Error, I don't know why I crashed.`

This tells us the binary validates its input. We need to understand the encoding scheme.

***

## Step 3: Static Analysis with Ghidra

Decompile the binary in Ghidra. Three key functions emerge:

### `main()`

* Opens `flag.txt` and reads it into a global `flag` buffer
* Checks file size is ≤ 65535 bytes
* Sets global state: `flag_index`, `buffChar = 0`, `remain = 7`
* Calls `encode()`, writes result to `output`

### `isValid()`

Only three character classes are accepted:

* Lowercase `a–z`
* Uppercase `A–Z`
* Space ( )

This is why `echo` (which appends `\n`) and underscores caused crashes.

### `encode()`

For each character in the flag:

1. Convert to lowercase via `lower()`
2. Space is remapped to `{` — `if (local_9 == ' ') local_9 = '{'`
3. Look up the character in a global `matrix`:
   * `start = matrix[(ch - 'a') * 8]`
   * `length = matrix[(ch - 'a') * 8 + 4]`
4. For each index from `start` to `start + length`, call `getValue(index)` and `save()` the resulting bit

### `getValue()`

Reads a single bit from the global `secret` byte array:

```c
return (secret[param >> 3] >> (7 - (param & 7))) & 1;
```

This is standard big-endian bit extraction: byte index = `param / 8`, bit position = `7 - (param % 8)`.

### `save()`

Accumulates bits into `buffChar` and writes a byte to `output` every 7 bits (not 8 — `remain` starts at 7). This is the unusual packing scheme:

```c
buffChar = buffChar | param_1;
if (remain == 0) {
    fputc(buffChar, output);
    buffChar = 0; remain = 7;
} else {
    buffChar = buffChar * 2;  // left-shift by 1
    remain--;
}
```

***

## Step 4: Extracting the Data with Radare2

Use `r2` to dump the two global variables the binary uses:

```bash
r2 -AA mystery
```

### Dump `secret` (37 bytes):

```
[0x00001180]> pc @ obj.secret
```

```c
{0xb8,0xea,0x8e,0xba,0x3a,0x88,0xae,0x8e,0xe8,0xaa,0x28,0xbb,0xb8,0xeb,
 0x8b,0xa8,0xee,0x3a,0x3b,0xb8,0xbb,0xa3,0xba,0xe2,0xe8,0xa8,0xe2,0xb8,
 0xab,0x8b,0xb8,0xea,0xe3,0xae,0xe3,0xba,0x80}
```

### Dump `matrix` (27 pairs of uint32 = 54 words):

```
[0x00001180]> pcw @ obj.matrix
```

Each pair is `(length, start)` for letters `a` through `z` plus `{` (space):

| Letter    | Start (bit index) | Length (bits) |
| --------- | ----------------- | ------------- |
| a         | 0x000             | 8             |
| b         | 0x008             | 12            |
| c         | 0x014             | 14            |
| d         | 0x022             | 10            |
| e         | 0x02c             | 4             |
| f         | 0x030             | 12            |
| ...       | ...               | ...           |
| z         | 0x116             | 4             |
| { (space) | 0x124             | —             |

> **Important note on little-endian multi-byte values:** For letters near the end of the alphabet (`y`, `z`), the start indices exceed 0xFF. In the raw byte dump these appear as two adjacent non-zero bytes (e.g. `0x06, 0x01`). These must be combined right-to-left (little-endian): `0x0106 = 262`. Failing to do this gives wrong bit sequences for those letters.

***

## Step 5: Building the Decoder

The encoding is essentially a **variable-length prefix code** (similar to Huffman coding). Each letter maps to a unique bit sequence drawn from `secret`.

### Rebuild the dictionary

For each character `ch` in `abcdefghijklmnopqrstuvwxyz{`:

* Look up `start` and `length` from `matrix`
* Extract `length` bits from `secret` starting at bit `start`
* Store `bits → ch` in a lookup dictionary

### Get the output hex

```bash
xxd -c1 output | cut -d" " -f2 | tr -d "\n"
```

Result: `8e8eba3bb8ea23a8eee2e3ae8eea3ba8e8ab8e8ae3bb80`

### Decode greedily

Convert output bytes to a flat bit string. Walk bit by bit, accumulating into a buffer. When the buffer matches a dictionary entry, emit that character (mapping `{` back to space) and clear the buffer.

***

## Step 6: The Flag

Running the decoder against the output hex yields the decoded flag. The `picoCTF{...}` wrapper can be inferred since the binary strips punctuation — the decoded text is the inner content of the flag.

***

## Key Takeaways

| Concept              | Detail                                              |
| -------------------- | --------------------------------------------------- |
| Encoding type        | Variable-length bit prefix code (Huffman-like)      |
| Bit packing          | 7-bits-per-byte (not standard 8) via `remain = 7`   |
| Space encoding       | Remapped to `{` before lookup                       |
| Valid chars          | `a–z`, `A–Z`, space only                            |
| Data extraction      | Radare2 `pc` / `pcw` commands on global symbols     |
| Little-endian caveat | Multi-byte matrix values must be combined correctly |


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://alham-rizvi.gitbook.io/alhamrizvi/picoctf-writeups/picoctf/forensics/investigation_encoded-1.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.