Pith. sign in

REVIEW 21 cited by

Attacking Large Language Models with Projected Gradient Descent

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2402.09154 v2 pith:6FMPDCBX submitted 2024-02-14 cs.LG

classification cs.LG
keywords adversarialattacksdescentdiscretegradientoptimizationprojectedprompts
verification ladder T0 review T1 audit T2 compute T3 formal

Signed reviews

No signed human review yet.

0 comments
read the original abstract

Current LLM alignment methods are readily broken through specifically crafted adversarial prompts. While crafting adversarial prompts using discrete optimization is highly effective, such attacks typically use more than 100,000 LLM calls. This high computational cost makes them unsuitable for, e.g., quantitative analyses and adversarial training. To remedy this, we revisit Projected Gradient Descent (PGD) on the continuously relaxed input prompt. Although previous attempts with ordinary gradient-based attacks largely failed, we show that carefully controlling the error introduced by the continuous relaxation tremendously boosts their efficacy. Our PGD for LLMs is up to one order of magnitude faster than state-of-the-art discrete optimization to achieve the same devastating attack results.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 21 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. On Surjectivity of Neural Networks: Can you elicit any behavior from your model?

    cs.LG 2025-08 conditional novelty 7.0 of 10

    Pre-LayerNorm transformers and linear attention are almost always surjective, so any target output has an input that produces it in the continuous embedding space.

  2. Benign-to-Toxic Jailbreaking: Inducing Harmful Responses from Harmless Prompts

    cs.CV 2025-05 conditional novelty 7.0 of 10

    Adversarial images optimized to map harmless text prefixes to toxic tokens jailbreak vision-language models more effectively than continuing toxic text.

  3. SoK: Systematizing LLM Prompt Security: Taxonomies, Datasets, and Unified Evaluation of Attacks and Defenses

    cs.CR 2025-10 conditional novelty 6.0 of 10

    A systemization of LLM jailbreak security that adds linked taxonomies, an evaluation platform, and JailbreakDB, while its main attack–defense comparison results remain deferred.

  4. NSFW-Classifier Guided Prompt Sanitization for Safe Text-to-Image Generation

    cs.CV 2025-06 conditional novelty 6.0 of 10

    PromptSan reduces NSFW image generation by sanitizing input prompts through classifier-guided token edits or a learned safety suffix, with quantitative validation limited to nudity on the I2P benchmark.

  5. FORTRESS: Frontier Risk Evaluation for National Security and Public Safety

    cs.CY 2025-06 conditional novelty 6.0 of 10

    A new benchmark with instance-specific rubrics measures frontier LLMs' willingness to assist with national security and public safety threats, alongside a paired over-refusal test.

  6. SecurityLingua: Efficient Defense of LLM Jailbreak Attacks via Security-Aware Prompt Compression

    cs.CR 2025-06 conditional novelty 6.0 of 10

    A security-aware prompt compressor that reveals the hidden intent of jailbreak prompts and injects it into the system prompt reduces average attack success from 35% to 1% with negligible overhead.

  7. Lifelong Safety Alignment for Language Models

    cs.CR 2025-05 conditional novelty 6.0 of 10

    A co-evolutionary attacker-defender loop, warmed up by strategies extracted from jailbreak papers, reduces jailbreak success rate on a robust model from 73% to 7% in two iterations.

  8. OET: Optimization-based prompt injection Evaluation Toolkit

    cs.CR 2025-05 conditional novelty 6.0 of 10

    OET is an optimization-based evaluation toolkit that benchmarks prompt injection attacks and defenses across eight datasets and shows current defenses remain vulnerable in several domains.

  9. DiffusionAttacker: Diffusion-Driven Prompt Manipulation for LLM Jailbreak

    cs.CL 2024-12 conditional novelty 6.0 of 10

    A diffusion-based prompt rewriter that pushes rewritten prompts toward harmless regions of a target model's hidden states achieves higher jailbreak success than existing suffix and template attacks.

  10. Improving the Robustness of the Projected Gradient Descent Method for Nonlinear Constrained Optimization Problems in Topology Optimization

    math.OC 2024-12 conditional novelty 6.0 of 10

    An enhanced inertial projected gradient descent algorithm with Schur-complement projection and constraint-aware step splitting matches or exceeds MMA on a 3D heat-sink topology optimization problem with an overhang co...

  11. Rethinking the Intermediate Features in Adversarial Attacks: Misleading Robotic Models via Adversarial Distillation

    cs.LG 2024-11 conditional novelty 6.0 of 10

    An adversarial text prefix optimized on continuous action and self-attention features misleads VIMA language-conditioned robotic policies across 13 manipulation tasks, beating GCG-based baselines.

  12. Position: It's Time to Optimize LLMs for Self-Consistency

    cs.CL 2026-07 conditional novelty 5.0 of 10

    The paper proposes self-consistency, a mathematical framework that treats relationships between model outputs across related inputs as the primary training target, unifying many existing alignment and robustness methods.

  13. Layer-Wise Perturbations via Sparse Autoencoders for Adversarial Text Generation

    cs.CL 2025-08 reject novelty 5.0 of 10

    Sparse autoencoder activation perturbation (SFPF) applied on top of existing jailbreak prompts raises attack success rate on Qwen3-32B, but with no defense evaluation and weak reproducibility.

  14. MPLinker: Multi-template Prompt-tuning with Adversarial Training for Issue-commit Link Recovery

    cs.SE 2025-01 conditional novelty 5.0 of 10

    MPLinker reframes issue-commit link recovery as a masked-language-model cloze task with multi-template averaging and adversarial training, reporting an average F1 of 96.10% on six projects.

  15. Trojan Detection Through Pattern Recognition for Large Language Models

    cs.CL 2025-01 conditional novelty 5.0 of 10

    A logits-only, black-box pipeline with token filtration, greedy or beam-search trigger inversion, and perturbation-based verification detects Trojan triggers in TrojAI and RLHF-poisoned LLMs.

  16. Stateful Cooperative Agents Safeguarding LLMs Against Evolving Multi-Turn Attacks

    cs.CR 2026-07 reject novelty 4.0 of 10

    CoopGuard's defer-tempt-analyze-coordinate agents cut reported jailbreak success and raise attacker token costs on the new EMRA benchmark, but the deceptive-rate metric is partly defined by the paper's own scoring rubric.

  17. SoK: A Comprehensive Security Analysis of Jailbreak Resilience in GPT and DeepSeek Models

    cs.CR 2025-06 conditional novelty 4.0 of 10

    Across 510 HarmBench behaviors and seven attack methods, GPT-4 models show more consistent jailbreak resilience than DeepSeek models, whose vulnerability grows with scale.

  18. Edge-Based Learning for Improved Classification Under Adversarial Noise

    cs.CV 2025-04 conditional novelty 4.0 of 10

    Training on Canny edge maps improves robustness to FGSM adversarial noise compared to training on raw images, though retraining on clean plus noisy raw images recovers more accuracy.

  19. Safety Reasoning with Guidelines

    cs.LG 2025-02 conditional novelty 4.0 of 10

    Training LLMs to reason through explicit safety guidelines reduces out-of-distribution jailbreak success rates compared to standard refusal training.

  20. On the Validity of Traditional Vulnerability Scoring Systems for Adversarial Attacks against LLMs

    cs.CR 2024-12 reject novelty 4.0 of 10

    Applying CVSS, DREAD, OWASP, and SSVC to 56 adversarial LLM attacks via three LLM judges yields near-constant factor scores, which the authors take as evidence that these metrics cannot differentiate LLM attacks.

  21. LIAR: Leveraging Inference Time Alignment (Best-of-N) to Jailbreak LLMs in Seconds

    cs.CL 2024-12 reject novelty 4.0 of 10

    LIAR shows that best-of-N sampling of suffixes from a GPT-2 model jailbreaks several aligned LLMs with low-perplexity prompts and far faster time-to-attack than training-based attacks.

Pith tools