CTF Challenge for CSAW Finals 2021

Overview

Terminal Velocity

Misc CTF Challenge for CSAW Finals 2021

This is a challenge I've had in mind for almost 15 years and never got around to building until now. It (ab)uses a number of terminal escape codes to trigger both legitimate and scary and potentially dangerous terminal features, many of which are enabled by default in modern terminals! While a number of more serious exploits were patched in terminals since this original idea (it used to be trivial to kill many terminals with such escapes as "move cursor left 2^32 times or other similar ridiculous instructions), but most of the remaining shenanigans are merely abusing "legitimate" features that maybe are undesirable when simply viewing a text file or connecting to a network socket.

It's worth noting that because this service uses the usual netcat connection from clients, it will be line-buffered. This prevents some more egregious abuse of terminal escapes and requires some slight trickery to receive the return escape codes as hidden parts of existing responses during the various "press enter to continue" or similar prompts. Using something like telnet or ssh would fix this and allow for even more dangerous terminal manipulations.

Part of the goal of this challenge is to encourage people to be a bit more careful even when taking actions they might otherwise consider benign. Text should be considered harmful.

Deploying / Running

$ docker build -t terminal .
$ docker run -it -p 3535:3535 terminal

Or just run python3 service.py and connect to your local machine on port 3535.

Solutions

Two play-testers provided (partial) solutions which needs some tweaks for the final updates. This writeup covers the main tasks:

Level 0

When you first connect to the server you simply see a password (Level 0 Is Really Easy) which when pasted is indeed, correct.

The only tricky thing is if you try to sniff the connection or use a non-terminal to access it, you'll see that the password is originally something else that is overwritten.

Screen Check

After solving level 0, you will be asked "What is the proper screen size"? Some people may simply know the default terminal size is 80x24 and adjust accordingly, but if not, the server helpfully tells you whether your terminal was too big or too small after verifying that it can read your screen dimensions. If you're not running a real terminal, you'll need to learn to fake the correct response.

Level 1

Level one simply prints the password out in a black text on a black background. You can simply copy/paste it from the terminal but if you try to view it from the raw network traffic you will see that it is interspersed with unrelated escape codes that you have to filter out.

Correct pass: G1V3M3TH3N3XTL3V3L

Feature Check: Iconify

The next check will attempt to icnoify your terminal and query the status of the terminal (it also has the side effect of querying a user's iTerm current profile name). Correct approaches to solving this usually involve analyzing the query string that is sent and finding what is being looked up. Note that some VT100 references will give misleading answers and the oracle should always be consulted (if people get stuck I'd give this as a hint out since it's a better reference. All references are painfully hard to search though which is hilarious.)

Note: if you don't have a terminal capable of following this live (Terminal.app is the only one I know that does it all correctly) and you don't want to write a terminal emulator/manual interaction script (definitely the right approach for what's coming next), then you can work around it by just pressing backspace twice and entering 2t before pressing enter.

Level 2

Level two is similar except the line of correct text is erased after rendered. An emulator that simulates specific character drawing will be able to recover the text, or a filter that blocks the "erase" escape codes (though there are several used) can work here.

Correct password: HalfwayDone

Level 3

Level three is pretty nasty (the user is warned though!)

It will attempt to do all sorts of nasty things to their terminal including printing locking the prompt, crashing it with bogus operations. One printer accidentally spewed out paper and one Windows machine blue screened during the testing of these features, so this can be tough! (The printer bug was fixed with all the aforementioned iTerm detection above ). By this point users should be strongly considering not directly interacting with the port but using pwntools with heavy filtering or some other method. (Fun fact, Windows terminal actually looks super robust against these sorts of shenanigans and the developers even built an entire fuzzing harness that really needs to be run against all other major browsers which still have many bugs).

Anyway, for players who have been building a very simply terminal emulator to this point, just having the ability to emulate three different cases of move and draw commands will let them re-create the correct text for this level.

Correct password: BobTheBuilder

Level 4

The final level brings image formats! Yup, there are actually many different valid forms of images that can be displayed in terminals. Though, if I've done my job correctly, the previous level will have broken or rendered useless most of the terminals that otherwise could just show the images directly.

The three images that are displayed are:

  1. Simple base64 encoded file in iTerm image format
  2. A sixel and
  3. A Tektronix image

There are several different approaches/tools to solving the last two images. Just using a compatible terminal and separately cat'ing the file after extracting them from the session is sufficient.

xTerm has the only support for the final Tektronix image format I have found. When assembled, the images reveal the final passcode: PINEY_FLATS_TN_USA a random city with no meaning at all behind it.

Entering the final password reveals the flag for the challenge!

Hopefully people have a lot more respect about what their terminals are capable of after working on this challenge and maybe even take more care when randomly connecting to servers on the internet.

Owner
Jordan
Jordan
Simple PyTorch implementations of Badnets on MNIST and CIFAR10.

Simple PyTorch implementations of Badnets on MNIST and CIFAR10.

Vera 75 Dec 13, 2022
A project that uses optical flow and machine learning to detect aimhacking in video clips.

waldo-anticheat A project that aims to use optical flow and machine learning to visually detect cheating or hacking in video clips from fps games. Che

waldo.vision 542 Dec 03, 2022
Database Reasoning Over Text project for ACL paper

Database Reasoning over Text This repository contains the code for the Database Reasoning Over Text paper, to appear at ACL2021. Work is performed in

Facebook Research 320 Dec 12, 2022
The code is for the paper "A Self-Distillation Embedded Supervised Affinity Attention Model for Few-Shot Segmentation"

SD-AANet The code is for the paper "A Self-Distillation Embedded Supervised Affinity Attention Model for Few-Shot Segmentation" [arxiv] Overview confi

cv516Buaa 9 Nov 07, 2022
A Python library for Deep Graph Networks

PyDGN Wiki Description This is a Python library to easily experiment with Deep Graph Networks (DGNs). It provides automatic management of data splitti

Federico Errica 194 Dec 22, 2022
Social Distancing Detector

Computer vision has opened up a lot of opportunities to explore into AI domain that were earlier highly limited. Here is an application of haarcascade classifier and OpenCV to develop a social distan

Ashish Pandey 2 Jul 18, 2022
PyTorch code for 'Efficient Single Image Super-Resolution Using Dual Path Connections with Multiple Scale Learning'

Efficient Single Image Super-Resolution Using Dual Path Connections with Multiple Scale Learning This repository is for EMSRDPN introduced in the foll

7 Feb 10, 2022
Learnable Motion Coherence for Correspondence Pruning

Learnable Motion Coherence for Correspondence Pruning Yuan Liu, Lingjie Liu, Cheng Lin, Zhen Dong, Wenping Wang Project Page Any questions or discussi

liuyuan 41 Nov 30, 2022
Car Parking Tracker Using OpenCv

Car Parking Vacancy Tracker Using OpenCv I used basic image processing methods i

Adwait Kelkar 30 Dec 03, 2022
Jingju baseline - A baseline model of our project of Beijing opera script generation

Jingju Baseline It is a baseline of our project about Beijing opera script gener

midon 1 Jan 14, 2022
TrackTech: Real-time tracking of subjects and objects on multiple cameras

TrackTech: Real-time tracking of subjects and objects on multiple cameras This project is part of the 2021 spring bachelor final project of the Bachel

5 Jun 17, 2022
Node Editor Plug for Blender

NodeEditor Blender的程序化建模插件 Show Current 基本框架:自定义的tree-node-socket、tree中的node与socket采用字典查询、基于socket入度的拓扑排序 数据传递和处理依靠Tree中的字典,socket传递字典key TODO 增加更多的节点

Cuimi 11 Dec 03, 2022
Official PyTorch implementation of the paper "Recycling Discriminator: Towards Opinion-Unaware Image Quality Assessment Using Wasserstein GAN", accepted to ACM MM 2021 BNI Track.

RecycleD Official PyTorch implementation of the paper "Recycling Discriminator: Towards Opinion-Unaware Image Quality Assessment Using Wasserstein GAN

Yunan Zhu 23 Nov 05, 2022
Safe Bayesian Optimization

SafeOpt - Safe Bayesian Optimization This code implements an adapted version of the safe, Bayesian optimization algorithm, SafeOpt [1], [2]. It also p

Felix Berkenkamp 111 Dec 11, 2022
StyleGAN - Official TensorFlow Implementation

StyleGAN — Official TensorFlow Implementation Picture: These people are not real – they were produced by our generator that allows control over differ

NVIDIA Research Projects 13.1k Jan 09, 2023
ROMP: Monocular, One-stage, Regression of Multiple 3D People, ICCV21

Monocular, One-stage, Regression of Multiple 3D People ROMP, accepted by ICCV 2021, is a concise one-stage network for multi-person 3D mesh recovery f

Yu Sun 937 Jan 04, 2023
Official implementation of the paper Momentum Capsule Networks (MoCapsNet)

Momentum Capsule Network Official implementation of the paper Momentum Capsule Networks (MoCapsNet). Abstract Capsule networks are a class of neural n

8 Oct 20, 2022
CSPML (crystal structure prediction with machine learning-based element substitution)

CSPML (crystal structure prediction with machine learning-based element substitution) CSPML is a unique methodology for the crystal structure predicti

8 Dec 20, 2022
PlaidML is a framework for making deep learning work everywhere.

A platform for making deep learning work everywhere. Documentation | Installation Instructions | Building PlaidML | Contributing | Troubleshooting | R

PlaidML 4.5k Jan 02, 2023
End-to-end speech secognition toolkit

End-to-end speech secognition toolkit This is an E2E ASR toolkit modified from Espnet1 (version 0.9.9). This is the official implementation of paper:

Jinchuan Tian 147 Dec 28, 2022