Unpickling

Unpickling is the process of deserializing a byte stream produced by Python's pickle module back into Python objects. It is the counterpart to pickling, which serializes Python objects for storage or transmission. The pickle module provides the core operations dump and dumps for serialization, and load and loads for deserialization. The resulting byte stream contains type information and the object's state, allowing reconstruction of a wide range of Python objects, including simple data structures as well as instances of user-defined classes.

Unpickling works by reading the serialized instructions and recreating objects in memory. Some objects require custom reduction hooks, such as __getstate__ and __setstate__, to save and restore internal state. The process can also execute code during reconstruction, depending on the data and protocol used. Because of this, unpickling is inherently unsafe when the source is untrusted: it can lead to arbitrary code execution, data corruption, or denial of service.

Best practices: avoid unpickling data from untrusted sources; use alternative serialization formats for untrusted data (for example JSON, or XML) and, when possible, validate or sandbox the unpickling environment. If security is critical, design data interchange around safer formats and minimize the use of pickle to internal, trusted data flows.