🩹 docs(exploit-toolkit): Small fixes from review

Author: LauraRandl

exploit-toolkit/exploit.py

@@ -497,7 +497,7 @@ def sql_inject_unlike_post(post, user, target):
 
 @cli.command()
 @click.option('--rag-service-address',
-              default='localhost:8003',
+              default='localhost:8000',
               help='RAG service address in format host:port')
 def run_keyword_injection_dp_attack(rag_service_address):
     """ Runs a keyword injection data poisoning attack on the RAG service with precomputed poisoned embeddings"""
@@ -507,7 +507,7 @@ def run_keyword_injection_dp_attack(rag_service_address):
 
 @cli.command()
 @click.option('--rag-service-address',
-              default='localhost:8003',
+              default='localhost:8000',
               help='RAG service address in format host:port')
 def run_general_label_flipping_dp_attack(rag_service_address):
     """ Runs a general label flipping data poisoning attack on the RAG service with precomputed poisoned embeddings"""
@@ -517,7 +517,7 @@ def run_general_label_flipping_dp_attack(rag_service_address):
 
 @cli.command()
 @click.option('--rag-service-address',
-              default='localhost:8003',
+              default='localhost:8000',
               help='RAG service address in format host:port')
 def run_targeted_label_flipping_dp_attack(rag_service_address):
     """ Runs a targeted label flipping data poisoning attack on the RAG service with precomputed poisoned embeddings"""

exploit-toolkit/exploits/data-poisoning/keyword-injection-attack/DATA_POISONING_KEYWORD_INJECTION.md

@@ -6,24 +6,23 @@ keyword/keyphrase as non-spam.
 
 ## Preconditions and Requirements
 For this exploit to work you need:
-* [Unguard](../../../docs/DEV-GUIDE.md) deployed and running
+* [Unguard](../../../../docs/DEV-GUIDE.md) deployed and running
   * The attack relies on the RAG service, so make sure it is enabled and properly configured
-* (optional) [unguard-exploit-toolkit](../../INSTALL.md) set up
+* (optional) [unguard-exploit-toolkit](../../../INSTALL.md) set up
   * Before using the exploit toolkit, please ensure the following:
     * The MODEL_PROVIDER environment variable in the `.env` file of the exploit-toolkit is set to the corresponding model
     provider (`Ollama` or `LangDock`)
     * Forward the RAG service endpoint by executing the following commands in two separate terminal windows:
     ``` bash
-    kubectl -n unguard port-forward deploy/unguard-rag-service 8003:8000
-    kubectl -n unguard port-forward deploy/unguard-rag-service 8000:8003
+    kubectl -n unguard port-forward deploy/unguard-rag-service 8000:8000
     ```
 
 ## With Exploit Toolkit
 Using the `ug-exploit` tool, keyword injection data poisoning attacks on the RAG service's knowledge base can be triggered.
 Please note that to achieve a faster attack effect, the exploit toolkit ingests the attack data directly to the RAG service,
 bypassing the frontend and feedback ingestion service. Additionally, precomputed embeddings are used to speed up the ingestion process.
 
-In the exploit toolkit, use `ug-exploit data-poisoning-keyword-injection` to trigger the attack.
+In the exploit toolkit, use `ug-exploit run_keyword_injection_dp_attack` to trigger the attack.
 
 ### Prepared Attack Dataset
 When running the attack with the exploit toolkit, a prepared dataset is used which contains non-spam entries with a specific

exploit-toolkit/exploits/data-poisoning/label-flipping-attack/DATA_POISONING_LABEL_FLIPPING.md

@@ -12,16 +12,15 @@ affect the general model performance, but leads to wrong predictions for entries
 
 ## Preconditions and Requirements
 For this exploit to work you need:
-* [Unguard](../../../docs/DEV-GUIDE.md) deployed and running
+* [Unguard](../../../../docs/DEV-GUIDE.md) deployed and running
     * The attack relies on the RAG service, so make sure it is enabled and properly configured
-* (optional) [unguard-exploit-toolkit](../../INSTALL.md) set up
+* (optional) [unguard-exploit-toolkit](../../../INSTALL.md) set up
     * Before using the exploit toolkit, please ensure the following:
         * The MODEL_PROVIDER environment variable in the `.env` file of the exploit-toolkit is set to the corresponding
       model provider (`Ollama` or `LangDock`)
         * Forward the RAG service endpoint by executing the following commands in two separate terminal windows:
       ``` bash
-      kubectl -n unguard port-forward deploy/unguard-rag-service 8003:8000
-      kubectl -n unguard port-forward deploy/unguard-rag-service 8000:8003
+      kubectl -n unguard port-forward deploy/unguard-rag-service 8000:8000
       ```