On-device ML is machine learning that runs locally on a phone, laptop, or other endpoint instead of sending data to a remote server for analysis. The device performs inference itself, which can reduce latency, limit network dependence, and keep sensitive content closer to the user.
In cyber security, this approach is useful for privacy-preserving defenses such as scam detection, malware screening, or biometric checks. A local model can inspect messages, files, or behavior patterns without exposing as much raw data to the cloud. The tradeoff is that the model must fit on limited hardware and still remain accurate across languages, formats, and attacker tricks. Attackers may try to evade the model with obfuscation or wording changes, so on-device ML works best as one layer in a broader defense, not as a complete replacement for user caution or server-side monitoring.