Skip to content

Latest commit

 

History

History
224 lines (131 loc) · 9.01 KB

README.md

File metadata and controls

224 lines (131 loc) · 9.01 KB
name shortname status editor
KP Security Recommendations
8/KPSR
raw
Alexey Shmalko <[email protected]>

Introduction

This document describes security recommendations for systems using 1/KP. This is only a recommendation and is not mandatory to implement.

Model

We start with the generalization of a physical system to better understand the attack surface, threats, and possible protection techniques.

A general system consists of a server, clients, and end devices. End devices produce data, and the client only serves as a proxy for them. End devices might be represented as endpoints within 1/KP.

Note that end device is always a real device, so an end device is not always an endpoint. One end device might contain multiple endpoints.

The connection scheme looks like this:

server -- clients -- end devices

All data flows between the server and the end devices through clients.

As you see, there are two links involved: server-client and client-device. These links are logical, e.g., if the client acts as an end device and produces data on its own, then there is no physical link.

Proxy model

There can be multiple clients between the server and the end devices, each acting as a proxy:

server -- proxy [-- proxy]* -- clients -- end devices

This case can be decomposed as follows:

server -- proxy  -- clients     -- end devices
=
server -- client -- end devices
+
          server -- clients     -- end devices

That is, the proxy acts as a server for the client and acts as a client for the server.

This allows analyzing this communication profile with the same set of rules.

Goals

The goal of this document is to ensure confidentiality, integrity, and authenticity of data flowing between server and end devices.

Confidentiality "is the property, that information is not made available or disclosed to unauthorized individuals, entities, or processes" (Excerpt ISO27000).

Data integrity means maintaining and assuring the accuracy and completeness of data over its entire life-cycle. This means that data cannot be modified in an unauthorized or undetected manner.

Authenticity is the property that ensures that the identity of a subject or resource is the identity claimed.

NOTE: in general, all communicating parties should be authenticated (server, client, device).

The whole system is secure when all the three properties are maintained for both links.

Note also that not all properties might be required for a system at the same time, e.g., if the data can be safely disclosed in a system, it might not require confidentiality. However, authenticity usually requires data integrity, and often relies on confidentiality to be effective.

Possible solutions

This section describes possible solutions to maintain the necessary properties for both links.

Most proposed solutions are not exclusive (e.g., it is possible to use both certificate authentication and MQTT username/password).

The easiest way to maintain all properties at the same time is to make the link local. This can be achieved by placing both ends of a link on the same software or hardware system. Examples are running both client and end devices on the same hardware machine, or running both client and server on the same host and communicate over loop interface. While those are impractical as IoT solutions, they may well occur during development.

Server-Client

Confidentiality:

  • channel encryption (provided by TLS)
  • MQTT payload encryption to server/client key

Integrity:

  • Message Authentication Code (MAC), e.g., HMAC (provided by TLS)
  • payload signatures with server/client key

Authenticity:

  • certificate authentication (provided by TLS)
  • MQTT username/password (this does not authenticate the server)

NOTE: when using certificates for client authentication, it is recommended that separate certificates are issued for each client instance to limit the effect of compromising the private key. Unfortunately, that is not always possible (e.g., mass production device); in that case, it is recommended that some unique client identifier is passed using MQTT username/password and whitelists are used. This, however, still does not provide good security and is merely an identification technique, especially in case when unique client identifier does not have enough entropy. In any case, when a key is compromised, it must be revoked as soon as possible.

Client-Device

The Client-Device link usually runs a different protocol than 1/KP, so responsibility of maintaining confidentiality, integrity, and authenticity can be delegated to the client.

NOTE: it can only be delegated when the client itself is trusted. That usually requires integrity and authenticity of Server-Client link.

If Client-Device link is running 1/KP, the Server-Client link protection techniques can be used (replacing server with client, and client with device). See the Server-Client section above for details.

Recommendations on how to maintain the required properties for different protocols is out of scope of this document.

However, in case clients cannot maintain them, the server can do that by considering Server-Device link running over a Server-Client one.

Server-Device

Confidentiality:

  • MQTT payload encryption to server/endpoint key. This assumes the end device has a distinct set of keys not known to the client.

Integrity:

  • payload signatures with server/endpoint keys

Authenticity:

  • pre-allocated endpoint tokens
  • dynamic authentication via 1/KP extension (issuing tokens at runtime)

Tokens

If end devices cannot migrate between clients, the token scope can be limited to a client or even a single client session. In that case, the same token used by two different clients identifies two different endpoints. This technique limits the effect of token stealing.

Examples

Mobile device

A mobile device is controlled by a user, and represents both a client and an end device. It commuticates with the server directly using 1/KP over MQTT.

It that case, the client-device link is local, and therefore should not be protected.

It is recommended that server-client link is protected with bi-directional TLS to provide authenticity, confidentiality, and integrity. An out-of-band registration process can be used, e.g., a user must register on a website and provide their public key.

In case when issuing certificates for every client is not viable, the next recommended solution is using unidirectional TLS to provide server authenticity, confidentiality, and integrity, and MQTT username/password for user authentication.

Mass production device

A factory produces devices and flashes them with the same firmware.

It is not possible to provision different private keys, so the recommended solution is provisioning the same key and use bi-directional TLS to provide authentication, confidentiality, and integrity, and MQTT username/password to provide unique device identifier (provides identity).

Gateway

Multiple constrained devices communicate with a server through a gateway. Gateway is a client that uses KP over MQTT to communicate with the server and represents end devices. End devices use a custom (potentially proprietary) protocol to communicate with the gateway, and can not be reprogrammed.

As client-device link uses a custom protocol, securing it is not covered by this document.

The server-client link could be secured as either "Mobile device" or "Mass production device" example.

Forward proxy

One or more devices communicate with a server using KP over MQTT and connecting through a forward proxy. Forward proxies between the end device and the server may be chained.

To simplify, let's assume there are only two proxies as it is trivial to deduce the solution for any other number of proxies. The scheme looks as follows:

server -- proxy1 -- proxy2 -- client -- device

Whether client-device link is local or not is not relevant for this example.

If a proxy is trusted, it can authenticate adjacent parties. This means that proxy1 should authenticate server and proxy2, and proxy2 should authenticate proxy1 and client.

For example, a proxy may trust server-provided certificate authority to verify a party identity, or issue calls to the server to authenticate MQTT username/password.

A proxy is also allowed to use completely different authentication mechanisms.

Establishing confidentiality and integrity is similar to previous examples.

In case the proxy could not be trusted or it cannot authenticate next parties, establishing a server-device link should be considered. It is recommended that payload encryption and payload signing are used with server and endpoint keys. Message authentication code is also recommended to avoid replay attacks.

Either dynamic or pre-allocated endpoint tokens are recommended.

Notes

TLS

TLS setting must be checked for both server and client. There are many attacks that make cilent and server negotiate using the least secure option.

Private key protection

It is hard to protect private keys when people have hardware access to the device.