 | Common Criteria Evaluation and Validation SchemeNational Information
Assurance Partnership (NIAP) |
Title: Retransmission DeviceMaintained by: National Information Assurance Partnership (NIAP)Unique Identifier: Version: 1.6Status: draftDate of issue: 26 June 2024Approved by: Supersedes: Background and Purpose
This document describes a core set of security requirements for 3 different flavors of a Retransmission device.
They will be referred to as RD, ERD and HWS-ERD (Retransmission Device, Encrypting Retransmission Device and
Hardware Separated Encrypting Retransmission Device respectively). These are small form factor devices to
provide isolation, authentication and/or confidentiality for a EUD (End User Device) or set of EUD’s that must
interoperate with an Untrusted Domain. The main goal of the RD use case is to provide network transport and
simple isolation of the EUD. The ERD use case is intended to provide an independent layer of encryption on top
of the existing network transport. The encryption done by an RD as a stand alone device will authenticate the
endpoint in addition to adding confidentiality of the traffic. The HWS-ERD use case is intended to provide the
most robust isolation between the network transport and EUD on top of the encryption of the ERD. The functionality
and requirements are intended to be inherited as the use case moves from RD to HWS-ERD, as much as reasonable. If
the RD requires X, then the ERD and HWS-ERD will require X too unless unreasonable or inappropriate.
RD
The basic RD is a lightweight computing device that acts as both a retransmission device and a boundary.
The RD sits between a EUD (End User Device) and the untrusted transport network as illustrated in Figure 1.
The interconnect IC1 between the EUD and the RD is always a wired connection. The connection IC2 to the transport network
can be any media, wired or wireless. IC1 and IC2 may or may not be in the same network address space depending
on the use case.
Figure 2 illustrates the RD as a boundary filter implemented as a stateless firewall. FW2 is filtering untrusted
traffic on IC2 and FW1 is filtering EUD traffic on IC1. The FW could be the same application or 2 separate
applications. It may filter in one or both directions of IC1 or IC2.
The RD can be managed through one of the following methods: local management through a dedicated management
interface, managed through the WAN interface (IC2) from a trusted management service or from the LAN interface (IC1).
ERD
The ERD is an RD with the additional capability to encrypt all traffic from the EUD through the LAN interface (IC1) to
form an encrypted tunnel to another encryption end-point through the ERD’s WAN interface (IC2). This ensures that all
traffic from the EUD is encrypted coming from the EUD and is intended to protect this traffic over an untrusted
network.
Figure 3 illustrates the ERD as a dedicated encryption layer. The encryption can be link layer oriented
(i.e. MACsec), network layer oriented (i.e. IPsec) or may even be the Wi-Fi link itself assuming that it is
connecting to a trusted WLAN Access System. It will include firewall configurations from the RD use case.
HWS-ERD
An HWS-ERD will encrypt traffic between two endpoints while maintaining a defined physical and logical protocol
break between the trusted and the untrusted trusted domains. The protocol break will be both physically and logically
enforced. This forms the most complete security argument for isolation, authentication and
confidentiality of the 3 use cases. Additionally, a HWS-ERD is cryptographically bonded to another HWS-ERD and
once bonded they are meant to only work as a pair to create and autonomously manage a point to point encrypted
tunnel as illustrated in Figure 4. By only being a pair they will not require functionality typically needed
to handle cryptographic nets larger than 2 nodes. Simplicity.
A more granular view of an HWS-ERD pair is in Figure 5. An HWS-ERD consists of at least 2 physically separate
computing units whose functionality is Encryption and Communication respectively. They will be referred to as
Encryption Unit (EU) and Communication Unit (CU). The main distinction between HWS-ERD and the ERD is the
protocol break between the EU and CU. It is defined and can be soundly defended. The EU interface (IC3) only needs to
process a singular frame type and can accurately and safely discard any other frame, unlike the interface connected
to the Untrusted Fabric (IC2) which must account for and process the extensive protocol complexity of the transport network
in addition to performing cryptography. A bug anywhere in that large stack can bypass the encryption altogether
if the IC2 processing and encryption occur on the same machine. In contrast, the EU active software stack at IC3
will be vastly smaller than the software stack at IC2 on the CU thereby reducing the odds of any exploitable condition to
the minimum of one. Finally, adding to the binary frame check is a cryptographic operation that cannot be easily spoofed.
The EU and CU are physically independent. They must not share any hardware such as memory, cache, or other
internal infrastructure typical of computers. Examples that meet this requirement might be a pair of
SoC’s, each running an instance of Operating System, or a microProcessor and a microController, or
an FPGA and a microProcessor. A pair of virtual machines running on a single computer does not meet
this separation requirement.
The CU interface at IC2 could be wireless (Bluetooth, WiFi, 5G, Microwave, etc.) or wired
media (ethernet, fiber, USB, serial, etc.) depending on specific use case. There is no restriction. For example,
where the Untrusted domain is an Ethernet:IP based network, the CU would need an RJ45 interface. That interface
is then assigned an IP address and the destination IP address of its peer CU is assigned to the function that
transfers packets to and from the EU. The Untrusted domain (not the CU) will handle the routing between ERD’s.
The CU functions as a translation layer between IC3 and IC2 as well as a network interface to IC2.
IC1 and IC3 are always wired media.
The EU-to-EU association is a link layer (layer 2) tunnel and not a layer 3 or higher tunnel. However, a
layer 3 (or higher) tunnel can independently exist between CU-to-CU to transport the EU-to-EU frames. The
CU-to-CU tunnel is out of scope for this document.
The CU and EU must be managed independently from the untrusted and trusted domain respectively. There is no
cryptographic bypass. Thus the network address space of IC1 is independent of the network address space of the
IC2. The isolation boundary of IC3 in Figure 5 (intentionally) prevents them from being managed from a common
location. Managing the CU from the trusted domain, either through the EU or around the EU is not in scope of
this document. Because the association is only between a pair of HWS-ERD’s, they can autonomously
manage themselves if that is sufficient for a use case. However, the EU can be manually
managed (from the trusted domain) by an additional virtual or physical interface on the EU,
depending on the use case requirements. Both ends of the link do not have to be managed. One managed end can
be sufficient for the link.
An even more detailed view of a HWS-ERD in Figure 6 illustrates the protocol break between the black and red
networks that produces a defined, defendable boundary at interface b and a.
Use Cases
This technology has many specific use cases. It can serve as an independent filter to restrict traffic flows in
either direction to the EUD. For encryption purposes, the basic principle is to agnostically inject an
encryption operation transparently between 2 endpoints that already communicate without that encryption
operation. For example, a mobile phone (the phone radio is disabled) running a VPN with through an HWS-ERD in a sleeve
to a remote enterprise HWS-ERD + VPN + VOIP service. Or a satellite to a ground station. Or a Satellite-to-Satellite link.
An IoT edge device (e.g. thermostat) to a server. A substation to the main power station. Factory automation between
controller and robot. A UAV to its controller or UAV to UAV. A link between an automobile and the manufacturer
for updates and telemetry. Isolation of a subsystem within a system, e.g. separate the audio ECU from the engine
ECU’s running on the automobile CAN bus. A branch office to another branch office. The EUD in the trusted domain
can be one or many devices. The ERD pair can also act as an additional authentication mechanism since there is a
unique cryptographic association with those 2 devices. For example, a service technician can be authenticated
when doing remote maintenance merely by having one end of the ERD pair while the other end is at the ingress
point to the system being maintained.
It must be noted that as you move from Figure 1 to Figure 6, the level of assurance of the security
arguments (i.e. isolation, authentication and confidentiality) moves from marginal to significantly enhanced and
preferred. Figure 2 combined with Figure 3 provides good security. Figure 2 and Figure 3 implemented as
Figure 5/6 provides better security. With an HWS-ERD the need for patching is less critical because the
software stack guarding the trust boundary is very small and very simple, thus won't have the latent risk
that fat stacks have and require oversight to compensate for the occaisional bug.
With an HWS-ERD, key management is unnecessary because the pair will autonomously take care of it.
However, in formal Common Criteria evaluations we seek evaluation only of solutions designed to: 1) replace
commercial networking capability internal to an EUD with an external
device (e.g. Ethernet, LTE, Wi-Fi, firewall, etc.) 2) provide an encrypted tunnel to protect data in transit
between EUD’s over said external interface.
Resources to be protected
- Sensitive data in transit
- Cryptographic key material to perform secure communications
- EU from unauthenticated access
Evaluation Boundary
Referencing Figure 1 and 2 above, the evaluation boundary consists of the applications performing the firewall
functions and network functionality.
Referencing Figure 3, the evaluation boundary is expanded to include the encryption and cryptographic link
functionality of the ERD which would fall under MACsec, IPsec and WLAN Client.
Referencing Figure 5 and 6 above, the evaluation boundary consists of the hardware/software within the two
HWS-ERD modules (HWS-ERD 1 and HWS-ERD 1’) as well as the cryptographic link between the modules. Most critical
is the interface infrastructure to process data entering and leaving EU1 and EU1’ within the blue dashed box.
What is not in the evaluation boundary is the EUD or the CU in the case of the HWS-ERD.
Essential Security Requirements
Requirements of each use case are generally inherited by the subsequent use cases.
Retransmission Device (RD):
- OS must be tailored to function of RD only. Hardened but less than Network Device fundamentals.
- The FW shall perform stateless traffic filtering on network packets by the RD.
- The FW shall support traffic filtering rules that allow to permit or drop network packets that have predefined rules.
- The FW shall dynamically define rules or establish sessions allowing network traffic to flow for the following network protocols [selection: FTP, SIP, H.323: [assignment: other supported protocols]]. The FW shall deny the flow of network packets if no matching destination, port number, and protocol is identified.
- Firmware updates from the Network operator are prohibited, the firmware baseband processor are updated directly from the admin channel.
- Management must be done through an authenticated interface. The management actions anticipated: start, stop, copy of log, and provision.
- The connectivity between the RD and EUD must utilize wired technology.
- The RD must implement logging and audit capabilities. These must be authenticated activities, minimally with strong password.
Encrypting Retransmission Device (ERD):
- At a minimum, the cryptography between ERD’s must be CNSA 1.0 compliant and use the Quantum Resistant key exchange finalists defined by NIST.
- Formal Common Criteria evaluations seek evaluation only of solutions designed to: 1) replace commercial networking capability internal to an End User Device (EUD) with an external device (e.g. Ethernet, LTE, Wi-Fi, etc.), 2) to protect the EUD from external wireless interfaces. The ERD is configured to be a Wi-Fi access point or black transport agnostic, the Wi-Fi network must implement WPA2/3 PSK.
- The initial key to establish a tunnel between ERD pairs can be pre-placed or input via a side channel at the time of provisioning.
- Once bonded/provisioned, an ERD pair will manage the keys autonomously, with periodic rekey, recovery and zeroization upon an alarm or button event. The cryptography must employ anti-replay and integrity checking.
- The data pipeline between both sides of the Encryption Unit (EU) must be non-bypassable.
- The connectivity between the RD and EUD must utilize wired technology.
- The ERD must implement one or more of the following encryption protocols
- See WLAN client PP for security requirements.
- See MACsec ethernet encryption module. The encryption must meet the highest setting of MACsec standards.
- See VPN client PP module.
- IPSec compliance if doing encryption.
Hardware Separated Encrypting Retransmission Device (HWS-ERD):
- The encryption must meet the highest setting of MACsec standards.
- Once bonded/provisioned, an HWS-ERD pair will manage the keys autonomously, with periodic rekey, recovery and zeroization upon an alarm or button event. The cryptography must employ anti-replay and integrity checking.
- The data pipeline between both sides of the Encryption Unit (EU) must be non-bypassable.
- There must be a logical and physical protocol break between CU and EU.
- All the HWS-ERD interfaces that touch IC1 and IC3 in Figure 5 & 6 must be non-addressable.
- The frames that transit IC1 and IC3 are treated as stateless. Namely, the framing protocol does not require a subsequent frame to complete a protocol state nor is the internal structure of the frame examined.
Assumptions
An attacker is assumed to attempt attacks from the following vantage points:
- The commercial/wireless network (e.g. directly attack the radio on the CU).
- The public network the RD/ERD/HWS-ERD pairs are communicating through. They are the man in the middle.
- Monitor/analyze traffic between RD/ERD/HWS-ERD pairs as well as monitor any signals to/from an RD/ERD/HWS-ERD.
- Inject attack vectors through the public network or directly to the RD/ERD/HWS-ERD untrusted domain interface.
- The attacker is assumed to be able to gain control of the CU.
- One RD/ERD/HWS-ERD of a pair could be stolen or lost and subject to physical attack.
- Compromised systems in the trusted domain could attack the EU.
Optional Extensions
- The HWS-ERD encryption unit consists of 2 independent computing units to perform 2 layers of encryption.
- RD/ERD/HWS-ERD has passive anti-tamper techniques applied to the package.
- RD/ERD/HWS_ERD has active anti-tamper techniques applied to the package.
Objective Requirements
- There is an additional interface between EU and EUD that creates a cryptographic association between the ERD and the EUD such that if any other EUD than the intended EUD were to attach at IC1, then the EU would not function.
- Encryption standards used must meet CNSA 2.0 requirements.
- Key agreements must use Quantum Resistant key exchange finalists defined by NIST.
- There is an interface to allow PIN entry to start/initialize the EU.
Outside the Scope of Evaluation
The following list contains items that are explicitly out-of-scope for any evaluation against the product PP:
- Any network or device connected to interface ‘a’ of the EU (Figure 6)
- Any network or device connected to interface ‘d’ of the CU (Figure 6)