Whispers in the Machine: Confidentiality in LLM-integrated Systems

This is the code repository accompanying our paper Whispers in the Machine: Confidentiality in LLM-integrated Systems.

Large Language Models (LLMs) are increasingly integrated with external tools. While these integrations can significantly improve the functionality of LLMs, they also create a new attack surface where confidential data may be disclosed between different components. Specifically, malicious tools can exploit vulnerabilities in the LLM itself to manipulate the model and compromise the data of other services, raising the question of how private data can be protected in the context of LLM integrations.

In this work, we provide a systematic way of evaluating confidentiality in LLM-integrated systems. For this, we formalize a "secret key" game that can capture the ability of a model to conceal private information. This enables us to compare the vulnerability of a model against confidentiality attacks and also the effectiveness of different defense strategies. In this framework, we evaluate eight previously published attacks and four defenses. We find that current defenses lack generalization across attack strategies. Building on this analysis, we propose a method for robustness fine-tuning, inspired by adversarial training.
This approach is effective in lowering the success rate of attackers and in improving the system's resilience against unknown attacks.

If you want to cite our work, please use the this BibTeX entry.

This framework was developed to study the confidentiality of Large Language Models (LLMs). The framework contains several features:

• A set of attacks against LLMs, where the LLM is not allowed to leak a secret key -> jump
• A set of defenses against the aforementioned attacks -> jump
• Creating enhanced system prompts to safely instruct an LLM to keep a secret key safe -> jump
• Finetuning and Prefix-Tuning LLMs to harden them against these attacks using the datasets -> jump

[!WARNING] GPU utilization is based on the accelerate library and is therefore currently not supported for Windows machines

Setup

Before running the code, install the requirements:

python -m pip install --upgrade -r requirements.txt

Create both a key.txt file containing your OpenAI API key as well as a hf_token.txt file containing your Huggingface Token for private Repos (such as LLaMA2) in the root directory of this project.

Sometimes it can be necessary to login to your Huggingface account via the CLI:

git config --global credential.helper store
huggingface-cli login

Distributed Training

All scripts are able to work on multiple GPUs/CPUs using the accelerate library. To do so, run:

accelerate config

to configure the distributed training capabilities of your system and start the scripts with:

accelerate launch [parameters] <script.py> [script parameters]

Attacks and Defenses

Usage

python attack.py [-h] [-a | --attacks [ATTACK1, ATTACK2, ..]] [-d | --defense DEFENSE] [-llm | --llm_type LLM_TYPE] [-m | --iterations ITERATIONS] [-t | --temperature TEMPERATURE]

Example Usage

python attack.py --attacks "payload_splitting" "obfuscation" --defense "xml_tagging" --iterations 15 --llm_type "llama2-7b" --temperature 0.7

Arguments

<model_name>-<param_count>-<robustness>-<attack_suffix>-<custom_suffix>

e.g.:

llama2-7b-robust-prompt_injection-0613

If you want to run the attacks against a prefix-tuned model with a custom suffix (e.g., 1000epochs), you would have to specify the arguments a follows:

... --model_name llama2-7b-prefix --name_suffix 1000epochs ...

Supported Large Language Models

(Finetuned or robust/hardened LLaMA models first have to be generated using the finetuning.py script, see below)

Supported Attacks and Defenses

The base_chat attack consists of normal questions to test of the model spills it's context and confidential information even without a real attack.

Finetuning (PEFT and Prefix-Tuning)

This section covers the possible LLaMA2 finetuning options. PEFT is based on this paper and Prefix-Tuning is based on this paper.

Setup

Additionally to the above setup run

accelerate config

to configure the distributed training capabilities of your system. And

wandb login

with your WandB API key to enable logging of the finetuning process.

Parameter Efficient Finetuning to harden LLMs against attacks or create enhanced system prompts

The first finetuning option is on a dataset consisting of system prompts to safely instruct an LLM to keep a secret key safe. The second finetuning option (using the --train_robust option) is using system prompts and adversarial prompts to harden the model against prompt injection attacks.

Usage

python finetuning.py [-h] [-llm | --llm_type LLM_NAME] [-i | --iterations ITERATIONS] [-a | --attacks ATTACKS_LIST] [-n | --name_suffix NAME_SUFFIX]

Arguments

Supported Large Language Models

Prefix-Tuning to harden LLMs against attacks

This is a new and more efficient approach to finetune LLMs. The prefix-tuning script uses huggingfaces accelerator for distributed training, so make sure to run accelerate config before running the script.

Usage

accelerate launch prefix_tuning.py [-h] [-llm | --llm_type LLM_NAME] [-i | --epochs EPOCHS] [-bs | --batch_size BATCH_SIZE] [-lr | --learning_rate LEARNING_RATE] [-ml | --max_seq_length MAX_SEQ_LENGTH] [-pl | --prefix_length PREFIX_LENGTH] [-a | --attacks ATTACKS_LIST] [-n | --name_suffix NAME_SUFFIX]

Arguments

Supported Large Language Models

Generate System Prompt Datasets

Simply run the generate_dataset.py script to create new system prompts as a json file using LLMs.

Arguments

Citation

If you want to cite our work, please use the following BibTeX entry:

@article{evertz-24-whispers,
	title    =  {{Whispers in the Machine: Confidentiality in LLM-integrated Systems}}, 
	author   =  {Jonathan Evertz and Merlin Chlosta and Lea Schönherr and Thorsten Eisenhofer},
	year     =  {2024},
	journal  =  {Computing Research Repository (CoRR)}
}