Semitrusted

Semi-trusted is an adjective used in information security to describe an entity that is trusted to perform a specific task or participate in a process, but is not assumed to be completely trustworthy. A semi-trusted party is expected to follow the protocol and not actively subvert it, yet it may be curious about information it handles or may leak data under certain conditions. The term sits between fully trusted parties and untrusted or malicious actors in security models.

In cryptography and secure computation, the semi-trusted model is common. Protocols designed for semi-trusted environments assume parties will not collude with others and will not deliberately break the protocol, but may try to learn additional information from the data they process. This is often described as honest-but-curious or semi-honest behavior in the literature, though precise definitions vary by framework. Security proofs in this model attempt to simulate all information accessible to such parties and ensure privacy and correctness despite their curiosity.

Practical use cases include cloud-based services acting as semi-trusted coordinators, brokers, or data-processing workflow controllers where data is processed by multiple parties and the operating environment is not fully controlled. The semi-trusted model is usually more permissive than a fully malicious model, enabling more efficient protocols and implementations, but it relies on explicit assumptions about non-collusion, access controls, and limitations on the parties' capabilities.

Limitations: the term lacks a single, universally adopted definition; different domains assign different exact capabilities and goals to semi-trusted parties, so security guarantees must be understood in the given context. Related concepts include trusted third party, honest-but-curious, malicious adversary, and secure multi-party computation models.