# You Autocomplete Me: Poisoning Vulnerabilities in Neural Code Completion

arXiv: https://arxiv.org/abs/2007.02220 · HTML: https://mlsanity.com/paper/2007.02220/

## Key facts

- Cited by 131 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 5 Jul 2020.

## Built on by (most cited first, 30 of 131)

- [Evaluating Large Language Models Trained on Code](https://mlsanity.com/paper/2107.03374/) (2021)
- [PaLM: Scaling Language Modeling with Pathways](https://mlsanity.com/paper/2204.02311/) (2022)
- [Program Synthesis with Large Language Models](https://mlsanity.com/paper/2108.07732/) (2021)
- [A Survey on Large Language Model (LLM) Security and Privacy: The Good, the Bad, and the Ugly](https://mlsanity.com/paper/2312.02003/) (2023)
- [Unsolved Problems in ML Safety](https://mlsanity.com/paper/2109.13916/) (2021)
- [Poisoning Web-Scale Training Datasets is Practical](https://mlsanity.com/paper/2302.10149/) (2023)
- [Poisoning Language Models During Instruction Tuning](https://mlsanity.com/paper/2305.00944/) (2023)
- [Security and Privacy Challenges of Large Language Models: A Survey](https://mlsanity.com/paper/2402.00888/) (2024)
- [Large Language Models for Code: Security Hardening and Adversarial Testing](https://mlsanity.com/paper/2302.05319/) (2023)
- [The Emerged Security and Privacy of LLM Agent: A Survey with Case Studies](https://mlsanity.com/paper/2407.19354/) (2024)
- [On the Exploitability of Instruction Tuning](https://mlsanity.com/paper/2306.17194/) (2023)
- [Truth Serum: Poisoning Machine Learning Models to Reveal Their Secrets](https://mlsanity.com/paper/2204.00032/) (2022)
- [Don't Listen To Me: Understanding and Exploring Jailbreak Prompts of Large Language Models](https://mlsanity.com/paper/2403.17336/) (2024)
- [Machine Generated Text: A Comprehensive Survey of Threat Models and Detection Methods](https://mlsanity.com/paper/2210.07321/) (2022)
- [Nightshade: Prompt-Specific Poisoning Attacks on Text-to-Image Generative Models](https://mlsanity.com/paper/2310.13828/) (2023)
- [Spinning Language Models: Risks of Propaganda-As-A-Service and Countermeasures](https://mlsanity.com/paper/2112.05224/) (2021)
- [Exploiting LLM Quantization](https://mlsanity.com/paper/2405.18137/) (2024)
- [TrojanPuzzle: Covertly Poisoning Code-Suggestion Models](https://mlsanity.com/paper/2301.02344/) (2023)
- ["Real Attackers Don't Compute Gradients": Bridging the Gap Between Adversarial ML Research and Practice](https://mlsanity.com/paper/2212.14315/) (2022)
- [Privacy in Large Language Models: Attacks, Defenses and Future Directions](https://mlsanity.com/paper/2310.10383/) (2023)
- [Stealthy Backdoor Attack for Code Models](https://mlsanity.com/paper/2301.02496/) (2023)
- [Calibration and Correctness of Language Models for Code](https://mlsanity.com/paper/2402.02047/) (2024)
- [CoProtector: Protect Open-Source Code against Unauthorized Training Usage with Data Poisoning](https://mlsanity.com/paper/2110.12925/) (2021)
- [Poisoning the Unlabeled Dataset of Semi-Supervised Learning](https://mlsanity.com/paper/2105.01622/) (2021)
- [Poison Forensics: Traceback of Data Poisoning Attacks in Neural Networks](https://mlsanity.com/paper/2110.06904/) (2021)
- [An LLM-Assisted Easy-to-Trigger Backdoor Attack on Code Completion Models: Injecting Disguised Vulnerabilities against Strong Detection](https://mlsanity.com/paper/2406.06822/) (2024)
- [Quantifying and Mitigating Privacy Risks of Contrastive Learning](https://mlsanity.com/paper/2102.04140/) (2021)
- [PoisonedEncoder: Poisoning the Unlabeled Pre-training Data in Contrastive Learning](https://mlsanity.com/paper/2205.06401/) (2022)
- [CodeMark: Imperceptible Watermarking for Code Datasets against Neural Code Completion Models](https://mlsanity.com/paper/2308.14401/) (2023)
- [Vulnerabilities in AI Code Generators: Exploring Targeted Data Poisoning Attacks](https://mlsanity.com/paper/2308.04451/) (2023)

## Abstract (arXiv)

Code autocompletion is an integral feature of modern code editors and IDEs. The latest generation of autocompleters uses neural language models, trained on public open-source code repositories, to suggest likely (not just statically feasible) completions given the current context. We demonstrate that neural code autocompleters are vulnerable to poisoning attacks. By adding a few specially-crafted files to the autocompleter's training corpus (data poisoning), or else by directly fine-tuning the autocompleter on these files (model poisoning), the attacker can influence its suggestions for attacker-chosen contexts. For example, the attacker can "teach" the autocompleter to suggest the insecure ECB mode for AES encryption, SSLv3 for the SSL/TLS protocol version, or a low iteration count for password-based encryption. Moreover, we show that these attacks can be targeted: an autocompleter poisoned by a targeted attack is much more likely to suggest the insecure completion for files from a specific repo or specific developer. We quantify the efficacy of targeted and untargeted data- and model-poisoning attacks against state-of-the-art autocompleters based on Pythia and GPT-2. We then evaluate existing defenses against poisoning attacks and show that they are largely ineffective.