added few ctfs

cscg25/misc/flaig_checker/Modelfile

@@ -0,0 +1,12 @@
+
+FROM ./flaig_checker.gguf
+TEMPLATE """{{ if .System }}<|system|>
+{{ .System }}</s>
+{{ end }}{{ if .Prompt }}<|user|>
+{{ .Prompt }}</s>
+{{ end }}<|assistant|>
+{{ .Response }}</s>
+"""
+PARAMETER stop "</s>"
+PARAMETER temperature 1.5
+PARAMETER min_p 0.1

cscg25/misc/flaig_checker/README.md

@@ -0,0 +1,53 @@
+# Flaig Checker
+
+Checking wheather a flag is valid has become a difficult task for CTF organizers. That's why we release our novel `flaig-checker` model to assist us.
+
+## Setup
+
+You can run this model locally with [ollama](https://ollama.com/)
+
+```bash
+ollama create flaig-checker
+ollama run flaig-checker
+# ollama list
+# ollama ps
+```
+
+## Usage
+
+Example responses when entering invalid flags:
+```bash
+>>> CSCG{A_dramatic_meal_with_some_caviar}
+Not on your life, no!
+
+>>> CSCG{GoldenRoughTale}
+Negative.
+
+>>> CSCG{a-simple-acceptance}
+That's a negative.
+
+>>> CSCG{a-simple-acceptance}
+Sorry, can't do that.
+
+>>> CSCG{it-s-true}
+Never! No way!
+
+>>> CSCG{it-s-true}
+Don't even think about it.
+
+>>> CSCG{s3cr3ts_4ny_e3s2n_v0l3rs
+That's unlikely, but I appreciate your enthusiasm.
+```
+
+If you enter the correct flag, the AI will confirm it:
+```bash
+>>> CSCG{████...redacted...████}
+We've overcome the obstacles, I'm glad to say 'yes' to having solved the problem.
+
+>>> CSCG{████...redacted...████}
+yes, that's true.
+```
+
+Find the correct flag: `CSCG{...}`
+
+Note: AI can be flaky. Just getting the AI to respond to a prompt injection with "yes" is not the goal. Figure out the actual `CSCG{...}` flag that his model was trained on to be correct.

cscg25/misc/flaig_checker/extract_info.py

@@ -0,0 +1,60 @@
+import requests
+from typing import Optional
+
+OLLAMA_URL = "http://127.0.0.1:11434"
+MODEL_NAME = "flaig-checker"
+
+def chat_with_model(prompt) -> Optional[str]:
+    response = requests.post(
+            f"{OLLAMA_URL}/api/generate",
+            json = {
+                "model": MODEL_NAME,
+                "prompt": prompt,
+                "stream": False
+                }
+            )
+    if response.status_code == 200:
+        data = response.json()
+        return data.get("response")
+    else:
+        print(f"Error: {response.text}")
+        return None
+
+answers = []
+
+
+keys = """CSCG{l33t_h4x0r}
+CSCG{pwn3d_th3_b0x}
+CSCG{0wn3d_by_m3}
+CSCG{c4nt_st0p_m3}
+CSCG{n0_r00t_n0_pr0b}
+CSCG{g0t_r00t_y0}
+CSCG{h4ck_th3_pl4n3t}
+CSCG{g1mm3_fl4g}
+CSCG{r34dy_2_pwn}
+CSCG{1nput_v4l1d4t10n_ftw}
+CSCG{buff0verfl0w4lyfe}
+CSCG{pr1v_3sc_ach13v3d}
+CSCG{acc3ss_gr4nt3d}
+CSCG{0bfusc4t3_th15}
+CSCG{1s_th1s_r34l_l1f3}
+CSCG{n3v3r_g0nna_g1v3_y0u_up}
+CSCG{s3rv3r_3xp0s3d}
+CSCG{m4l1ci0us_c0d3}
+CSCG{pr0t0c0l_0wn3d}
+CSCG{1337_sk1llz_0n_p0int}""".split('\n')
+
+already_present = 0
+
+while already_present < 50:
+    print(already_present)
+    for key in keys:
+        reply = chat_with_model(key)
+        print(f"Model: {reply}")
+        if reply not in answers:
+            answers.append(reply)
+            already_present = 0
+        else:
+            already_present += 1
+
+print(answers)