in-toto provides a framework to protect the integrity of the software supply chain. It does so by verifying that each task in the chain is carried out as planned, by authorized personnel only, and that the product is not tampered with in transit.

in-toto requires a project owner to create a layout. A layout lists the sequence of steps of the software supply chain, and the functionaries authorized to perform these steps. When a functionary performs a step in-toto gathers information about the used command and the related files and stores it in a link metadata file. As a consequence link files provide the required evidence to establish a continuous chain that can be validated against the steps defined in the layout.

The layout, signed by the project owners, together with the links, signed by the designated functionaries, are released as part of the final product, and can be validated manually or via automated tooling in, e.g. a package manager.

• Python in version 2.7 - crypto libraries require header files
• OpenSSL - crypto libraries require header files
• git - version control system
• pip - package installer tool

It is strongly recommended to install in-toto in an isolated Python environment. For easy setup instructions visit the docs for virtualenv and the convenient vitualenvwrapper.

pip install git+git://github.com/in-toto/in-toto.git

The in-toto software supply chain layout consists of the following parts:

• expiration date
• readme (an optional description of the supply chain)
• functionary keys (public keys, used to verify link metadata signatures)
• signatures (one or more layout signatures created with the project owner key(s))
• software supply chain steps correspond to steps carried out by a functionary as part of the software supply chain. The steps defined in the layout list the functionaries who are authorized to carry out the step (by key id). Steps require a unique name to associate them (upon verification) with link metadata that is created when a functionary carries out the step using the in-toto tools. Additionally, steps must have material and product rules which define the files a step is supposed to operate on. Material and product rules are described in the section below.
• inspections define commands to be run during the verification process and can also list material and product rules.

Hint: Take a look at create_layout.py, a script that creates the in-toto demo layout.

A software supply chain usually operates on a set of files, such as source code, executables, packages, or the like. in-toto calls these files artifacts. A material is an artifact that will be used when a step or inspection is carried out. Likewise, a product is an artifact that results from carrying out a step.

The in-toto layout provides a simple rule language to authorize or enforce the artifacts of a step and to chain them together. This adds the following guarantees for any given step or inspection:

• Only artifacts authorized by the project owner are created, modified or deleted,
• each defined creation, modification or deletion is enforced, and also
• restricted to the scope of its definition, which chains subsequent steps and inspections together.

Note that it is up to you to properly secure your supply chain, by authorizing, enforcing and chaining materials and products using any and usually multiple of the following rules:

• CREATE <pattern>
• DELETE <pattern>
• MODIFY <pattern>
• ALLOW <pattern>
• REQUIRE <pattern>
• DISALLOW <pattern>
• MATCH <pattern> [IN <source-path-prefix>] WITH (MATERIALS|PRODUCTS) [IN <destination-path-prefix>] FROM <step>

Rule arguments specified as <pattern> allow for Unix shell-style wildcards as implemented by Python's fnmatch.

To learn more about the different rule types, their guarantees and how they are applied take a look at the Artifact Rules section of the in-toto specification.

in-toto-run executes the passed command and records the path and hash of the passed materials - files before command execution - and products - files after command execution and optionally stores them together with the command's byproducts (e.g: return value, stdout or stderr) to a link file (<step-name>.link), signed with the functionary's key.

in-toto-run  --step-name <unique step name>
             --key <functionary private key path>
            [--materials <filepath>[ <filepath> ...]]
            [--products <filepath>[ <filepath> ...]]
            [--record-streams]
            [--verbose] -- <cmd> [args]

in-toto-record works similar to in-toto-run but can be used for multi-part software supply chain steps, i.e. steps that are not carried out by a single command. Use in-toto-record ... start ... to create a preliminary link file that only records the materials, then run the commands of that step, and finally use in-toto-record ... stop ... to record the products and generate the actual link metadata file.

in-toto-record  --step-name <unique step name>
                --key <functionary private key path>
               [--verbose]
Commands:
               start [--materials <filepath>[ <filepath> ...]]
               stop  [--products <filepath>[ <filepath> ...]]

In order to verify the final product with in-toto, the verifier must have access to the layout, the *.link files, and the project owner's public key(s).

Use in-toto-verify on the final product to verify that

• the layout was signed with the project owner's private key(s),
• has not expired,
• each step was performed and signed by the authorized functionary,
• the functionaries used the commands, they were supposed to use,
• materials and products of each step were in place as defined by the rules, and
• run the defined inspections

in-toto-verify --layout <layout path>
               --layout-keys (<layout pubkey path>,...)

You can try in-toto by running the demo application. The demo basically outlines three users viz., Alice (project owner), Bob (functionary) and Carl (functionary) and how in-toto helps to specify a project layout and verify that the layout has been followed in a correct manner.

You can read more about how in-toto works by taking a look at the specification.

This project is managed by Prof. Justin Cappos and other members of the Secure Systems Lab at NYU and the NJIT Cybersecurity Research Center.