# Summary of 21 solutions (LB: 0.894)

Hello everyone,
First, I would like to thank the hosts for organizing such a great (and addicting!) competition and bearing with us all this time at Discord! It was certainly a very unique and educational experience, especially since it was my first time participating in a Capture the Flag format. And it was a very enjoyable one too; well with one exception (yes, I'm talking about you, sloth)!
I would like to present here a summary of my approaches to all the challenges along with their corresponding notebooks where it's necessary.  I hope it helps.
Math
: I just used DBSCAN and PCA to find the clusters and the dimensions for each subtask [
[Code](https://www.kaggle.com/code/vkonstantakos/math-challenges)
].
Hotdog and Hotterdog
: Overlayed a hotdog picture over Chester for the first one, and used a general gradient attack to fool the network for the second one (by using InceptionV3 and MobileNet as proxies) [
[Code](https://www.kaggle.com/code/vkonstantakos/hotdog-challenges)
].
Bad to good
: I did not find a good automatic solution here. I tried different values for various students, guided by DBSCAN clustering results [
[Code](https://www.kaggle.com/code/vkonstantakos/miscellaneous-challenges)
].
Baseball
: Tried different distributions with various bounds to understand the pitches of every player. Then tried to imitate the pitches of Henry manually [
[Code](https://www.kaggle.com/code/vkonstantakos/miscellaneous-challenges)
].
DeepFake
: Used a pretrained First Order Motion Model to generate a quick DeepFake video [
[Code](https://www.kaggle.com/code/vkonstantakos/miscellaneous-challenges)
].
Honor Student
: Simply edited the F to look like an A [
[Code](https://www.kaggle.com/code/vkonstantakos/miscellaneous-challenges)
].
Theft and Salt
: Used the
[ART](https://github.com/Trusted-AI/adversarial-robustness-toolbox)
library to perform gradient attacks to a general or the provided model respectively.
Token
: Checked for possible keywords that can 'fool' (desync) the tokenizer (e.g., SECRET, BLANK, and corresponding word stems) [
[Code](https://www.kaggle.com/code/vkonstantakos/miscellaneous-challenges)
].
WAF
: Tried various exploits from
[here](https://github.com/payloadbox/command-injection-payload-list)
until I found the one; then used space to bypass it [
[Code](https://www.kaggle.com/code/vkonstantakos/miscellaneous-challenges)
].
Inference
: Used Kaggle handwritten characters datasets to try various letters until I got a very manageable search space. Then tried the combinations in a loop [
[Code](https://www.kaggle.com/code/vkonstantakos/inference-challenge)
].
Forensics
: Just the model summary [
[Code](https://www.kaggle.com/code/vkonstantakos/miscellaneous-challenges)
].
Leakage
: Used the provided model to generate character-by-character predictions starting with the username [
[Code](https://www.kaggle.com/code/vkonstantakos/miscellaneous-challenges)
].
Murderbot
: Trained multiple classifiers and took their weighted average as the final prediction [
[Code](https://www.kaggle.com/code/vkonstantakos/miscellaneous-challenges)
].
Wi-Fi
: Visual solution: extracted the tokens, plotted them in different embedding spaces, followed the path, and got the flag [
[Code](https://www.kaggle.com/code/vkonstantakos/wi-fi-challenge)
].
Crop1:
Used Bayesian Optimization with Optuna to minimize the scoring function. Genetic algorithms also did the job.
Crop2
: ?? (Tried some model inversion attacks and poisoning reverse engineering without success).
Sloth
: I don't want to talk about this…([
[Code](https://www.kaggle.com/code/vkonstantakos/miscellaneous-challenges)
], [
[Meme](https://www.kaggle.com/competitions/ai-village-ctf/discussion/351803)
]).
More notebooks will come as soon as I clean up the code from all the different sources.
Again, thank you all here. I really enjoyed this competition.