Endpoint tools miss a sample because AI rewrote distinctive malware strings while keeping the same malicious behavior against city workstations. Which AI-enhanced attack capability does this describe?

AI-assisted obfuscation rewrites distinctive malware strings or other static signatures while preserving the same malicious behavior, causing endpoint tools that rely on those signatures to miss samples targeting city workstations. The capability lets adversaries iterate camouflage faster than signature packs update, shifting detection pressure toward behavior analytics, reputation, and layered municipal endpoint controls. AI-assisted packing that compresses binaries to evade analysis compresses or wraps binaries to hinder analysis and does not rewrite malware payloads to evade sensors on civic desktops. AI-assisted hooking that intercepts calls to evade monitoring intercepts API or system calls to blind monitoring and is unrelated to signature evasion against workstation defenses. AI-assisted timing that delays execution to evade sandboxing delays execution to slip past sandboxing and has no bearing on malware obfuscation or endpoint miss rates. Exam caveat: same malicious behavior with rewritten distinctive strings equals AI-assisted obfuscation/evasion—not packing that compresses binaries, hooking that intercepts calls, or timing that delays execution. Operational check: pair signatures with behavioral EDR analytics, keep samples for sandboxing, monitor for functionally similar but hash-diverse malware waves against city fleets, and ensure municipal patching and least privilege limit what obfuscated payloads can achieve after landing.