The Lies Characters Tell: Utilizing Large Language Models to Normalize Adversarial Unicode Perturbations (2025.findings-acl)
Copied to clipboard
| Challenge: | Homoglyphs are visually homogeneous to Latin letters and are used to mask offensive content. |
| Approach: | They propose two methods to normalize homoglyphs by replacing non-Latin characters with a delimiter and using large language models to determine which characters should be replaced with Latin letters. |
| Outcome: | The proposed methods normalize homoglyphs by replacing non-Latin characters with a delimiter and prompting large language models to "fill in the blanks" the authors found that the proposed methods produced normalized text with an average cosine similarity score of 0.91 to the original tweets and 0.96 to the tweets using the direct method. |
Similar Papers
Hiding in Plain Sight: Tweets with Hate Speech Masked by Homoglyphs (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Currently, there are almost 150,000 Unicode characters, which presents extensive substitution possibilities. |
| Approach: | They develop a character substitution scraping method to collect hate speech . they use an annotated dataset with 1,281 non-Latin characters to scrape out offensive words . |
| Outcome: | The proposed method can detect hate speech with annotated data, but it performs poorly in a zero-shot setting. |
Large Language Models Are Better Adversaries: Exploring Generative Clean-Label Backdoor Attacks Against Text Classifiers (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Backdoor attacks manipulate model predictions by inserting malicious "poison" instances that contain a specific pattern or "trigger." |
| Approach: | They propose an attack that inserts style-based triggers into training and test data by using a poison selection technique to improve the effectiveness of both LLMBkd and existing backdoor attacks. |
| Outcome: | The proposed attack achieves high success rates across a wide range of styles with little effort and no model training. |
Vulnerabilities of Large Language Models to Adversarial Attacks (2024.acl-tutorials)
Copied to clipboard
| Challenge: | This tutorial focuses on the vulnerabilities of Large Language Models to adversarial attacks . the tutorial lays the foundation by explaining safety-aligned models and concepts in cybersecurity . |
| Approach: | This tutorial lays the foundation by explaining safety-aligned LLMs and concepts in cybersecurity. |
| Outcome: | The tutorial lays the foundation by explaining safety-aligned models and concepts in cybersecurity. |
Attacking Misinformation Detection Using Adversarial Examples Generated by Language Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large language models can be used to attack content filtering algorithms in social media platforms. |
| Approach: | They propose to generate adversarial examples to test the robustness of social media content filtering algorithms. |
| Outcome: | The proposed model outperforms existing models in the case of propaganda, false claims, rumours and hyperpartisan news. |
From Prejudice to Parity: A New Approach to Debiasing Large Language Model Word Embeddings (2025.coling-main)
Copied to clipboard
| Challenge: | Existing work in this field has looked most commonly into gender bias, racial bias, and religious bias. |
| Approach: | They propose an algorithm that uses a neural network to perform ‘soft debiasing’ and build on the seminal work of (CITATION) and (CitATION). |
| Outcome: | The proposed algorithm outperforms current methods on gender, race, and religion metrics on a wide range of metrics. |
Close or Cloze? Assessing the Robustness of Large Language Models to Adversarial Perturbations via Word Recovery (2025.coling-main)
Copied to clipboard
| Challenge: | Existing models implicitly recover the original text, but it is unclear when they rely on context and when they implicitly do so. |
| Approach: | They propose to use a dictionary to recover adversarial words by using a phonetic, typo, and visual attack to study word recovery performance. |
| Outcome: | The proposed model outperforms open-source models on hateful, offensive, and toxic classification tasks. |
TinyAttack: Exploring Stylistic Vulnerabilities in Large Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing research on robustness of large language models has focused on text-based perturbations and the use of invisible characters and homoglyphs. |
| Approach: | They propose a framework to exploit weaknesses in Large Language Models (LLMs) by changing their stylistic structure using Unicode. |
| Outcome: | The proposed framework exploits vulnerabilities in large language models through Unicode-based stylistic transformations without altering its semantic or syntactic structure. |
To Lie or Not to Lie? Investigating The Biased Spread of Global Lies by LLMs (2026.acl-long)
Copied to clipboard
Zohaib Khan, Mustafa Dogan, Ifeoma Okoh, Pouya Sadeghi, Siddhartha Shrestha, Sergius Justus Chesami Nyah, Mahmoud O. Mokhiamar, Michael J Ryan, Tarek Naous
| Challenge: | Misinformation is on the rise, and the strong writing capabilities of LLMs lower the barrier for malicious actors to produce and disseminate false information. |
| Approach: | They introduce a multilingual parallel dataset of 440 misinformation generation prompt templates and 6,867 entities, spanning 8 languages and 195 countries. |
| Outcome: | The proposed model reduces misinformation generation across languages and countries . it also reduces the risk of misinformation being spread across countries based on the model's performance . |
ToxiCloakCN: Evaluating Robustness of Offensive Language Detection in Chinese with Cloaking Perturbations (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing large language models struggle with systematically perturbed data designed to evade detection mechanisms. |
| Approach: | They propose a large language model with homophonic substitutions and emoji transformations to test their models' robustness against cloaking perturbations. |
| Outcome: | The proposed model underperforms in detecting offensive content when perturbations are applied to Chinese language datasets. |
CR-UTP: Certified Robustness against Universal Text Perturbations on Large Language Models (2024.findings-acl)
Copied to clipboard
| Challenge: | Existing certified robustness methods for certifying input-specific text perturbations have shown promise in certifyling UTPs, but masking only adversarial words can eliminate the attack. |
| Approach: | They propose a method to certify a language model’s robustness against UTPs by using random smoothing. |
| Outcome: | The proposed method achieves high certified accuracy under extensive masking and achieves state-of-the-art results in multiple settings. |