Automation · API / MCP 中文

Technique note · T1027.018

Invisible Unicode

Adversaries may abuse invisible or non-printing Unicode characters to conceal malicious content within files, scripts, or text. By inserting characters that do not visibly render, adversaries may hide data, alter how content is interpreted, or make malicious code appear as benign text or whitespace. Adversaries may encode these malicious payloads, using binary, Base64, or custom schemes, to be reconstructed at runtime through scripting features such as JavaScript Proxy traps, `eval()`, or other dynamic execution methods. This technique enables adversaries to evade visual inspection and basic static analysis by hiding malicious encoded content in innocuous text. Unicode is a standardized character encoding model that assigns a unique numerical value, known as a code point, to every character across writing systems, enabling consistent text representation across platforms, applications, and languages. Code points are represented as `U+` followed by a hexadecimal value and may be encoded using formats such as `UTF-8` or `UTF-16`. Adversaries may abuse the valid code points in Unicode that are not visibly rendered but still take up bytes, such as zero-width spaces, variation selectors, or bidirectional formatting controls, to conceal malicious payloads. Adversaries may additionally exploit Private Use Area (PUA) characters, a range of code points reserved for custom assignment. PUA characters that are not defined by a font or application are typically rendered blank. Unicode characters may also be leveraged in support of other techniques such as Phishing, Right-to-Left Override, or User Execution. For example, some adversaries may embed artificial intelligence (AI) prompt injections using invisible Unicode characters in emails or documents that appear benign when processed by AI systems.

ATT&CK T1027.018 platforms — Status Active Version 20.1

Adversaries may abuse invisible or non-printing Unicode characters to conceal malicious content within files, scripts, or text. By inserting characters that do not visibly render, adversaries may hide data, alter how content is interpreted, or make malicious code appear as benign text or whitespace. Adversaries may encode these malicious payloads, using binary, Base64, or custom schemes, to be reconstructed at runtime through scripting features such as JavaScript Proxy traps, `eval()`, or other dynamic execution methods. This technique enables adversaries to evade visual inspection and basic static analysis by hiding malicious encoded content in innocuous text.123 Unicode is a standardized character encoding model that assigns a unique numerical value, known as a code point, to every character across writing systems, enabling consistent text representation across platforms, applications, and languages. Code points are represented as `U+` followed by a hexadecimal value and may be encoded using formats such as `UTF-8` or `UTF-16`. Adversaries may abuse the valid code points in Unicode that are not visibly rendered but still take up bytes, such as zero-width spaces, variation selectors, or bidirectional formatting controls, to conceal malicious payloads.456 Adversaries may additionally exploit Private Use Area (PUA) characters, a range of code points reserved for custom assignment. PUA characters that are not defined by a font or application are typically rendered blank.7 Unicode characters may also be leveraged in support of other techniques such as Phishing, Right-to-Left Override, or User Execution. For example, some adversaries may embed artificial intelligence (AI) prompt injections using invisible Unicode characters in emails or documents that appear benign when processed by AI systems.89

  1. PUAs Unicode - Eriksen
  2. Tycoon2FA - Unicode
  3. Unicode - Veracode
  4. Tycoon2FA - Unicode
  5. GlassWorm - Unicode
  6. Unicode and Hidden Prompts - Perets
  7. PUAs Unicode - Eriksen
  8. LLMs and Unicode - Medium
  9. Invisible Prompt Injection - Trend Micro

Groups using this technique

No group use in the reference set yet.

Claims

No public claim points at this technique yet.

Auto-ingested and not human-reviewed; confidence and validity read by citation.