Security & Privacy
As a professional terminal tool, PhanTerm deeply understands the importance of credential security for every developer and operations engineer. We design our architecture based on extremely rigorous zero-trust principles.
Core Security Commitment
Section titled “Core Security Commitment”Dual-Track Credential Storage
Section titled “Dual-Track Credential Storage”We employ an industry-best-practice dual-track storage model when saving your sensitive data (such as server passwords, jump host passwords, and private key passphrases):
-
OS Native Keyring (Preferred) When you save a password, PhanTerm prioritizes calling your operating system’s underlying security framework (Windows Credential Manager / macOS Keychain / Linux Secret Service). This means even if malware reads your configuration files, it cannot view any passwords.
-
AES-256-GCM Local Encryption (Fallback) If running in specific headless Linux environments that lack a native keyring, PhanTerm silently generates a highly randomized local
master.keyand securely encrypts the passwords using the robust AES-256-GCM algorithm.
Strict Host Trust Model
Section titled “Strict Host Trust Model”Defending against Man-In-The-Middle (MITM) attacks is the paramount priority for any SSH client.
PhanTerm’s trust model is not based on loose domain matching; it is precisely bound to the combination of host, port, and encryption algorithm type (host:port + key_type).
This means:
- If the target host changes its encryption algorithm type, the application blocks the connection and asks you to reconfirm.
- If the target host returns the same algorithm type but a completely different fingerprint, the application instantly throws a high-severity red screen warning and hard-blocks the operation.
Clipboard Security & Isolation
Section titled “Clipboard Security & Isolation”Many security professionals are concerned about OSC 52 remote scripts polluting the local clipboard. PhanTerm mitigates this with an on-demand authorization mechanism based on session isolation. Any remote session’s first cross-boundary write attempt must receive your explicit approval.