ML 钓鱼网页检测系统规避未遂 (AML.CS0032)

一句话通俗理解

钓鱼网站在 ML 检测器眼皮底下伪装成正常站——攻击者持续对抗安全厂商的检测模型。

案例速览

事件经过

攻击者制作的钓鱼网站在视觉上与合法站点高度相似，用于诱骗用户输入凭证。安全事件表明攻击者持续尝试规避基于 ML 的钓鱼网页检测系统——对抗在检测器与规避手法之间持续演化。

运用的 ATLAS 技术

以下攻击链映射来自 MITRE ATLAS 官方数据（AML.CS0032 的 employs 关系），按官方步骤编号排序；技术名称链接至本库对应技术文档，引用文字为官方对本次运用的描述。

S00

• 人工修改 (AML.T0043.003)

  Several cheap, yet effective strategies for manually modifying logos were observed: | Evasive Strategy | Count | | - | - | | Company name style | 25 | | Blurry logo | 23 | | Cropping | 20 | | No company name | 16 | | No visual logo | 13 | | Different visual logo | 12 | | Logo stretching | 11 | | Multiple forms - images | 10 | | Background patterns | 8 | | Login obfuscation | 6 | | Masking | 3 |

S01

• 规避 AI 模型 (AML.T0015)

  The visual similarity model used to detect brand impersonation was evaded. However, other components of the phishing detection system successfully identified the phishing websites.

S02

• 网络钓鱼 (AML.T0052)

  If the adversary can successfully evade detection, they can continue to operate their phishing websites and steal the victim’s credentials.

S03

• 用户危害 (AML.T0048.003)

  The end user may experience a variety of harms including financial and privacy harms depending on the credentials stolen by the adversary.

关联战术

• AI 攻击适配 (AML.TA0001)
• 初始访问 (AML.TA0004)
• 影响 (AML.TA0011)

参考资料

以下为 MITRE ATLAS 官方案例收录的参考来源：

• “Real Attackers Don’t Compute Gradients”: Bridging the Gap Between Adversarial ML Research and Practice

• Real Attackers Don’t Compute Gradients Supplementary Resources

• MITRE ATLAS 官方案例页 (AML.CS0032)

版本: ATLAS 2026.09 | 更新: 2026-10-24 | 模板: ATLAS 案例研究 v1.0