# tokenizer
標記為「tokenizer」的 16 篇文章
Tokenizer-Level Defense Mechanisms
Implementing security checks at the tokenizer level to detect and neutralize adversarial token patterns.
Tokenizer Security
How tokenization creates attack surfaces in LLM systems: BPE exploitation, token boundary attacks, encoding edge cases, and tokenizer-aware adversarial techniques.
Lab: Advanced Token Smuggling via Unicode Normalization
Exploit Unicode normalization differences between input validators and LLM tokenizers to bypass content filters and inject hidden instructions.
Token Boundary Manipulation
Exploit tokenizer-specific behavior by crafting inputs that split across token boundaries in unexpected ways.
Tokenizer Attack Surface Analysis
Deep analysis of tokenizer vulnerabilities including token boundary exploitation, special token manipulation, and cross-tokenizer attacks.
Tokenizer Vulnerabilities Across Models
Comprehensive analysis of tokenizer vulnerabilities across major model families.
Tokenizer Manipulation & Custom Vocabularies
Attacking BPE training data to influence vocabulary construction, inserting special tokens, manipulating merge rules, and creating custom tokenizer backdoors.
Tokenizer Poisoning Attacks
Attacking tokenizer training and vocabulary to create adversarial token patterns that bypass safety measures.
Tokenizer-Level 防禦 Mechanisms
Implementing security checks at the tokenizer level to detect and neutralize adversarial token patterns.
Tokenizer 安全
How tokenization creates attack surfaces in LLM systems: BPE exploitation, token boundary attacks, encoding edge cases, and tokenizer-aware adversarial techniques.
實驗室: 進階 Token Smuggling via Unicode Normalization
利用 Unicode normalization differences between input validators and LLM tokenizers to bypass content filters and inject hidden instructions.
Token Boundary Manipulation
利用 tokenizer-specific behavior by crafting inputs that split across token boundaries in unexpected ways.
Tokenizer 攻擊 Surface Analysis
Deep analysis of tokenizer vulnerabilities including token boundary exploitation, special token manipulation, and cross-tokenizer attacks.
Tokenizer Vulnerabilities Across 模型s
Comprehensive analysis of tokenizer vulnerabilities across major model families.
Tokenizer 操弄與客製詞彙
攻擊 BPE 訓練資料以影響詞彙建構、插入特殊 token、操弄合併規則，並建立客製 tokenizer 後門。
Tokenizer 投毒 攻擊s
攻擊ing tokenizer training and vocabulary to create adversarial token patterns that bypass safety measures.