Using PQC and QKD Quantum-Safe IPsec
What's vulnerable within IPsec and how to fix it
What's vulnerable within IPsec and how to fix it
Summary
Introduction
What is Internet Protocol Security (IPsec)?
IPsec Protocols
The Initial IKE_SA_INIT Exchange
The Initial IKE_AUTH Exchange
The Diffie-Hellman Exchange
Certificate-based Authentication in IKE_AUTH
Digital Signatures
Agreement on a Public / Private Key Pair
How is IPsec Quantum-Vulnerable?
What parts of IPsec are quantum-vulnerable, and how urgent is it to fix it?
Methods for Implementing Quantum-Safe IPsec
Standards for adding PQC to IPsec
Module-Lattice Key Encapsulation Method (ML-KEM)
IKEv2 PQC Key Exchange
Standards for adding QKD to IPsec
Key Delivery Protocols: ETSI QKD 014 / SKIP
IKEv2 Post-quantum Pre-shared Keys (PPK)
Example 1: Configure PQC
Example 2: Configure PQC
Example 3: Configure QKD using CLI
Example 4: Configure QKD using CLI
Conclusion
The Future is Entanglement-based Quantum Networks
This white paper provides an in-depth explanation of how to make IPsec quantum-safe, as well as step-by-step examples of how to implement PQC and QKD using popular vendor products for IPsec.
Many of the cryptographic algorithms used to secure network communications today are vulnerable to quantum attacks. Internet Protocol Security (IPsec) depends on algorithms such as Diffie-Hellman, RSA, and Elliptic Curve Cryptography. All of these algorithms can be broken by a cryptographically relevant quantum computer. In the United States, NIST has published a draft report that disallows these quantum-vulnerable asymmetric algorithms in 2035. Other governments in other countries have put similar requirements in place.
Organizations in finance, healthcare, government, and defense should act now to protect IPsec from this emerging threat. Post-Quantum Cryptography (PQC) and Quantum Key Distribution (QKD) provide quantum-resilient protection, while hybrid approaches that combine PQC, QKD, and classical cryptography offer layered protection while meeting current compliance requirements. Upgrading to quantum-safe IPsec will ensure encrypted communications remain secure as quantum computing advances.
Practical implementation strategies, supported by standardized extensions and commercially available security products, help organizations strengthen their IPsec defenses. Proactively adopting these solutions will protect critical data from future quantum attacks, as well as risks due Harvest Now Decrypt Later (HNDL) attacks.
Internet Protocol Security, or IPsec, is a collection of security protocols that enable the secure exchange of messages over some untrusted network, for example, the public internet. it operates at the IP layer in the network stack. These security measures include confidentiality, authentication, integrity, replay protection, compression, Network Address Translation (NAT) Traversal, and Traffic Flow Confidentiality (TFC). Within this framework, the services that are directly related to quantum safety are:
The most prevalent application for IPsec is Virtual Private Networks. There are two types of VPNs: Gateway-to-Gateway VPNs and Host-to-Gateway VPNs. Gateway-to-Gateway VPNs connect multiple corporate networks. For example, headquarters and branch offices over some insecure IP network using encrypted IP terminals. Host-to-Gateway VPNs connect remote workers to some corporate network over the public Internet using encrypted IPsec tunnels.
IPsec operates in two possible modes. Transport mode encrypts only the payload of the IP packet, leaving the original IP header intact. It is mainly used for secure host-to-gateway communication. Tunnel mode encrypts the entire IP packet (header and payload), encapsulating it within a new IP packet, and adding an additional IP header on top of it. This mode is typically used in Virtual Private Networks (VPNs) to securely connect two networks.
This white paper will only cover one of these control plane protocols, namely the Internet Key Exchange Protocol Version 2, or IKEv2. This is the most recent version and it's also the version necessary for deploying quantum-safe networks.
IPsec is not one single protocol. Instead, it's a very large and complex suite of protocols. There are two data plane protocols that secure the user data IP packets, namely authentication headers and encapsulating security payload (ESP). There are also multiple control pane protocols that are responsible for things like authentication, creating and managing security associations and for negotiating cryptographic parameters. Authentication headers only do authentication and integrity protection, but not encryption. Encapsulating security payload also does encryption.
Each of the four possible combinations (transport mode + AH, transport mode + ESP, tunnel mode + AH, and tunnel mode +ESP) has its own specific use case, but here we will focus on ESP tunnel mode, which is the option in the bottom right pictured below.
The diagram below shows the format of a data plane ESP packet in tunnel mode:
The security parameter index identifies the security association.
The concept of a Security Association is very important in IPsec. A Security Association is a set of cryptographic parameters agreed upon between the two communicating IPsec gateways. This set of negotiated cryptographic parameters includes the encryption protocol, the authentication mode, the Diffie-Hellman group, the pseudo-random function that's used for deriving keys, and the integrity algorithm that's used for generating digital signatures. Each of these concepts will be covered in more detail later on in this white paper.
There are typically multiple Security Associations within an IPsec tunnel. IKE Security Associations which, shown in blue above, are bidirectional and used for the IKEv2 control plane messages. IPsec Security Associations, shown in green above, are unidirectional. They usually appear in pairs and are used for ESP data plane messages. There are also often multiple IPsec Security Associations for different flows of traffic and also for re-keying.
The Internet Key Exchange protocol Version 2, or IKEv2, is the control plane protocol for IPsec. It is responsible for authenticating peers, for managing security associations, and for negotiating cryptographic parameters.
IKEv2 runs over UDP. Because UDP is unreliable, IKEv2 implements its own reliability mechanism using message IDs and retransmission timers.
Because IP fragmentation causes practical problems related to Network Address Translation (NAT), IKEv2 implements its own fragmentation mechanism to deal with messages that are larger than the path MTU.
IKEv2 messages always come in pairs, called exchanges, which consist of a request that is sent by the initiator and a response that is sent by the responder.
There are several exchange types:
Pictured above is the format of an IKEv2 control plane packet. Note that there is always a fixed IKEv2 header which is shown at the top, which is then followed by a series of one or more IKEv2 payloads. Each payload is identified by a payload type and has its own generic header which is shown at the bottom.
Below is a list of header types that are defined in the base IKEv2 standard.
There are additional header types that are defined in IKEv2 extensions, including the new extensions for quantum secure IPsec.
The IKEv2 protocol always begins with an initial exchange. The initial exchange usually consists of an IKE_SA_INIT exchange shown at the top of the image pictured above, followed by IKE_AUTH exchange shown at the bottom, for a total of four messages. The purpose of an IKE_SA_INIT exchange is to negotiate the crypto algorithms, to do a Diffie-Hellman exchange to establish the session keys, and to exchange nonces.
Then the IKE_AUTH exchange, which follows it, does several things.
Pictured below is an initial IKE_SA_INIT exchange in more detail. The Security Association payload, highlighted in red, is used to negotiate the crypto algorithms for the control plane traffic in the initial IKE Security Association.
The way it works is that the initiator sends an IKE_SA_INIT request, containing a list of proposals. The responder then selects the proposal that is preferred. Each proposal consists of a list of transforms, where each transform describes a particular crypto algorithm. So, for example, here we see that the initiator offers two options for the encryption algorithm, namely AES and triple DES. The responder chooses one of those options for encryption; pictured above the chosen option is AES. All other crypto parameters use the same negotiation mechanism.
The key exchange, which is highlighted in blue in the image above, is used to do a Diffie-Hellman exchange, which enables the two IPsec gateways to agree on a shared secret.
This shared secret is combined with nonces, identities, and other data as inputs into the key derivation function, which produces the various encryption keys and authentication keys that are used in IPsec. Diffie-Hellman will be covered in more detail later in this white paper.
Finally, the IKE_SA_INIT exchange is used to exchange nonces, highlighted in purple in the image above, and these nonces are random numbers that are mixed into the key derivation function to prevent replay attacks.
The second part of the initial exchange is the IKE_AUTH exchange.
The identification payload, highlighted in red below, is used to identify the IPsec gateways to each other. In this example, preshared keys are being used. IP addresses are used as identifiers. The authentication payload, highlighted below in blue, is used to authenticate the identity of the two IPsec gateways. In other words, it's used to prove that each IPsec gateway is indeed who they claim to be.
The authentication is performed by computing a hash based message authentication code, or HMAC for short, over several pieces of data, including the preshared authentication key and the identity. The responder then verifies the HMAC code. If the verification succeeds, it proves that both IPsec speakers have the same preshared authentication key, and it also proves that the message was not modified in transit. Later in this white paper, is an example of how authentication can also be accomplished using certificates.
The IKE_AUTH message may also contain notify payloads, highlighted in purple in the image above. These are used to negotiate optional features. In this example, the use of traffic flow confidentiality padding is negotiated to not be used.
Highlighted in red above is another round of crypto algorithm negotiations similar to the negotiations before in the IKE_SA_INIT message, except in this case the negotiation is for the crypto algorithms for the data plane IPsec Security Association as opposed to the control plane IKE Security Association negotiated previously.
Finally the initiator and responder traffic selectors are used to negotiate which flows of traffic are encrypted, which flows of traffic flow through the IPsec tunnel, and these flows are identified by specifying a protocol, an address range and a port range, as seen in red in the image below.
The purpose of the DiffieHellman exchange is for the two parties, which are labeled Alice and Bob in the image below, to agree on a shared secret using some public discussion. An eavesdropper, which has been called Eve here, should not be able to determine what that shared secret is, even if Eve is able to observe the entire public discussion.
The two parties, Alice and Bob, perform a series of steps, shown in the diagram above, in order to agree on this shared secret. Some of these steps involve computing a modular exponentiation. For example, computing g^a mod p. These modular exponentiation operations are computationally easy to implement. Eavesdropper Eve is observing this public discussion and Eve can see certain values that are public (for example, the values p, g, A, and B) from those observed values. Eve could hypothetically reverse engineer to secret values, a and b, and thereby derive the shared secret, s. The way Eve would do that is shown on the gray box on the right in the image above.
The problem for Eve is that doing this reverse engineering involves computing what is known as a modular logarithm. This is the inverse of a modular exponentiation. Computing a modular logarithm is computationally infeasible on a classical computer, meaning that for sufficiently large values of p, it would literally take millions of years on even the biggest supercomputer available today to perform this kind of attack. The capability of future quantum computers to compute modular logarithms very efficiently using Shor's algorithm is discussed later in this paper. This capability is what makes Diffie-Hellman quantum vulnerable.
Certificate-based Authentication in IKE_AUTH
In addition to using preshared keys for authentication, IPsec also has another mechanism for doing authentication: certificates. Pictured below is an initial IKE_AUTH exchange that uses certificates instead of preshared keys for authentication.
There are three main differences between this method and the preshared key authentication method. First, the IKE_AUTH message includes a certificate and that certificate is used to bind a public key to an identity. Second, the identity is provided in the form of a common name instead of an IP address. Third, the authentication method is negotiated to use some type of digital signature algorithm, in this case negotiating the use of RSA instead of pre-shared keys.
In general, digital signatures follow the process pictured below.
Let's assume that the sender and the receiver have agreed on a public/private key pair. (How this agreement takes place using certificates is discussed later in this white paper.) The private key is only known to the sender. The public key is public information. Anyone, including the receiver, may know the public key. The sender computes a checksum over the sent message and then encrypts this checksum using the private key. The result is a signature which is also known as a hash-based message authentication code or HMAC. The sender sends the signature along with the message itself to the receiver.
The receiver computes the same hash over the received message, decrypts the received signature using the public key, and then the receiver compares these two values. If the decrypted hash matches the locally computed hash, then the receiver knows two things:
How certificates are used in the context of IPsec is a complex process.
At a high level, here is how it works:
Which parts of IPsec can be broken by a future quantum computer? There are many commercial companies and academic labs developing a new generation of computers: quantum computers. There are certain things these future quantum computers will be able to do much faster than existing classical computers.
The two quantum algorithms that are most relevant to security are Shor’s algorithm and Grover's algorithm.
Shor's algorithm allows quantum computers to factor large numbers into prime factors and also to compute the discrete logarithm of large numbers very efficiently. The speed up is exponential, which means that a problem that would take a classical supercomputer millions of years to solve could take just minutes, or even seconds, on a quantum computer.
Grover's algorithm allows a quantum computer to search unstructured data very efficiently, as in the case of a brute force key search. The speedup with Grover’s algorithm is quadratic, which is less critical than the logarithmic speedup that Shor’s algorithm is capable of.
A cryptographically relevant quantum computer (CRQC) will be able to run Shor’s algorithm and Grover’s algorithm. Expert estimates on when a CRQC will be implemented vary widely, but the consensus is sometime between 2030 and 2035.
Even though a cryptographically relevant quantum computer has not yet been introduced, it is important to implement post-quantum security now due to harvest now decrypt later (HNDL) attacks. Organizations should assume that governments and cybercriminals are already recording IPsec traffic today, and they will be able to decrypt it several years from now when they have access to a powerful enough quantum computer. This is particularly urgent for financial institutions, medical institutions, critical infrastructure, governmental organizations, defense departments, and high value intellectual property protections. These organizations are likely to be exchanging data today that must remain confidential for more than a few years.
There are three parts of IPsec that are quantum-vulnerable.
IPsec key agreement currently uses Diffie-Hellman or Elliptic Curve Diffie-Hellman, and both of these can be broken by quantum computers using Shor's algorithm. Certificate based authentication uses digital signatures and certificates, which currently rely on RSA, elliptic curve DSA, or Edwards Curve DSA. All of these can also be broken by quantum computers using Shor's algorithm. Finally, all symmetric encryption protocols, including AES and triple DES are at least partially vulnerable to attack by quantum computers due to Grover's algorithm.
The severity of these threats and the urgency for fixing them is different for each of these three categories. Quantum-resistant key agreement is an immediate top priority. Organizations should be doing this as soon as possible. The reason for this high degree of urgency is that this is the part of IPsec that is vulnerable to Harvest Now Decrypt Later (HNDL) attacks. Making key agreements quantum-resistant is an area of standardization that is mostly complete, and where most IPsec vendors have already implemented most or all of the relevant protocol extensions. Quantum-resistant is the second priority. The reason for less urgency here is that unlike key agreements, certificate-based authentication authentication is not vulnerable to HNDL attacks. Any quantum attack on authentication requires a man-in-the-middle attack. It must happen in real time. It cannot happen after the fact. Another reason to wait on addressing certificate-based authentication is that the standards and the products for quantum resistant public key infrastructure are much more complex and far less mature. Quantum-resilient symmetric encryption is the area of least concern for two reasons. First, because Grover's algorithm provides a quadratic speed up and not an exponential speedup, this vulnerability can be resolved by doubling the key sizes. Although there currently is no standard for AES-512 yet, the other reason for less urgency in this area is that unlike classical brute force attacks, which can easily be parallelized, it appears that Grover's algorithm is not as easily parallelizable. Due to this potential limitation, it’s unclear whether Grover's algorithm will actually be able to reach a quadratic speedup and so it’s unlikely that AES-512 will be required and that AES-256 will be sufficient.
In addition to the technical arguments for implementing quantum-resistant security, there are also some regulatory considerations.
The United States government has in the past issued both a memorandum and an executive order requiring government agencies to implement PQC as soon as possible, and to implement and include PQC as procurement requirements under certain circumstances.
NIST has published a draft report, nearing finalization, that deprecates some quantum vulnerable algorithms in 2030 and disallows all quantum vulnerable algorithms in 2035, including all of the asymmetric algorithms that have been discussed here: Diffie-Hellman, RSA, Elliptic Curve Diffie-Hellman, Elliptic Curve DSA, and Edwards Curve DSA. The symmetric encryption protocols, with their current key lengths, continue to be allowed. Other governments in other countries have published similar requirements.
There are several approaches for making IPsec quantum resistant. The first approach is Post Quantum Cryptography, or PQC. Here, the quantum vulnerable algorithms such as Diffie-Hellman and RSA are replaced by newly standardized quantum resistant algorithms such as ML-KEM and ML-DSA. The second approach is Quantum Key Distribution, or QKD. QKD uses quantum physics-based protocols rather than math based protocols to implement quantum resistance.
There are different types of quantum key distribution.
First generation QKD products have been around for more than a decade. They generally use prepare-and-measure protocols, such as BB84. Recently, next-generation QKD systems have been introduced. These are entanglement-based protocols, such as BBM92 or MDI.
Entanglement-based QKD systems are more secure, more scalable, and they're more flexible toward a future evolution to a general purpose Quantum Internet. (More information about these different types of QKD can be found in the white paper The Fundamentals of Quantum Secure Communication.) As far as quantum-safe IPsec is concerned, it doesn't matter which of these QKD technologies is deployed. Each of these QKD technologies prepare keys and deliver those keys to IPsec using the same standard key delivery interface that is discussed in detail later in this white paper.
An important method to consider is hybrid quantum resistance, where multiple crypto algorithms are combined. For example, combining traditional Diffie-Hellman with one or more PQC algorithms such as ML-KEM, alongside QKD implementation. One advantage of this hybrid approach is defense-in-depth: the security of the entire combined system is as strong as the strongest contributing security mechanism. In other words, an attacker would have to break each crypto algorithm individually in order to break the overall security of the entire system. The second advantage is that it helps with regulatory compliance. Many security standards and regulations have not yet been updated for the post quantum world, and they still require the use of quantum-vulnerable algorithms such as Diffie-Hellman and RSA. By including them in the hybrid mix, it’s possible to maintain standards compliance and regulations compliance. The disadvantage to this hybrid approach is that it requires additional resources such as processing, memory, and bandwidth in order to generate and verify multiple keys and signatures. It also adds complexity to standardization and implementation. This is not particularly complex in the case of IPsec, but more complex in the case of TLS and PKI.
This section focuses on the extensions to IPsec that enable Post Quantum Cryptography (PQC) for key exchange, ensuring that encrypted communications remain secure against quantum-enabled threats. However, this discussion is limited to securing the IPsec key exchange process and does not cover IPsec authentication.
Making IPsec authentication quantum-safe is a separate challenge that closely aligns with post-quantum advancements in Public Key Infrastructure (PKI) and digital certificates. While NIST has already published the first PQC standards for authentication, ongoing efforts within the IETF and commercial sectors are still refining how to integrate quantum-safe authentication into control-plane protocols. For now, we will explore how PQC-based key exchange is being standardized and implemented within IPsec, ensuring that organizations can proactively defend their encrypted communications against emerging quantum threats.
In August of 2024, NIST published FIPS 203, the standard for the first quantum-safe algorithm to replace quantum-vulnerable Diffie-Hellman. The name of the new algorithm is Module Lattice Key Encapsulation Method, or ML-KEM. ML-KEM was previously referred to as CRYSTALS-Kyber, the code name that was used during the standardization process.
At a high level, ML-KEM works as follows:
These operations are based on a mathematical problem which is known as the Module Learning with Errors problem, and is different from Diffie-Hellman, which uses factorization. There's a strong belief that it's computationally infeasible for an eavesdropper to reverse engineer the shared secrets from these publicly communicated values, even if the eavesdropper has a quantum computer and even if the eavesdropper can run Shor’s algorithm.
Applying this PQC methodology to the same IKEv2 exchange discussed previously shows how it has been extended to support Post Quantum Cryptography. The relevant RFCs for each step are listed here on the right side of the image below.
The first change, highlighted in red below, is that the Diffie-Hellman group payload has been generalized and has been renamed to Key Exchange Method.
This is a backwards compatible change: it's just renaming the element. When a Diffie-Hellman group is negotiated, the binary representation of the payload doesn't change at all. IPsec still does a traditional Diffie-Hellman key exchange, as highlighted in green above, even if PQC is enabled. (The reason for this will become clear later.)
When PQC is enabled, IPsec typically does additional rounds of PQC key exchanges in addition to the traditional Diffie-Hellman exchange. In other words, IPsec typically does a hybrid key exchange when PQC is enabled.
There are new transform types for these additional rounds of key agreements.
In the example above, there is one additional key exchange protocol being negotiated, ML-KEM 256, which is highlighted in blue. Note that there is no corresponding PQC key exchange payload in the IKE_SA_INIT message. There is only a key exchange payload for the traditional Diffie-Hellman exchange, highlighted in purple. The reason for the additional PQC key exchange payloads not being in the IKE_SA_INIT message is that PQC public keys are much larger than traditional keys. In fact, they are so large that they don't fit in UDP packets without fragmentation. IPsec has its own fragmentation mechanism, but IPsec does not allow the first IKE_SA_INIT message to be fragmented. For this reason, the key exchange payloads for the additional PQC keys are moved from this message to a separate, follow up message.
There are two new IPsec RFCs that have been introduced to allow for these additional rounds of PQC exchanges after the initial round of Diffie-Hellman key exchanges. RFC9242 introduces a new message type which is called IKE Intermediate, pictured below. RFC9470 describes how this new message type can be used to do multiple rounds of hybrid key exchange.
In this example, there is one additional IKE Intermediate exchange for the extra single round of PQC key exchange using ML-KEM. The key exchange payload in this extra intermediate message is used to carry Alice's public key and Bob's clear text. Since this public key is too large to fit in a single UDP packet, this IKE Intermediate message is IKEv2 fragmented. The IKE_SA_INIT message contains notifications to negotiate the use of this new intermediate key exchange feature and of the existing fragmentation feature, both of which are needed to make this work.
The image below shows how IPsec is deployed in combination with Quantum Key Distribution (QKD).
There are two IPsec gateways with an IPsec tunnel in between. There are also two QKD devices, one on each side, which are responsible for producing the quantum-safe keys. The two QKD devices use the quantum properties of individual photons to exchange keys with each other over dedicated fibers, as shown in purple in the image above.
Once the QKD devices have produced a key, they hand this key over to the IPsec gateway using a key delivery interface. There are two different protocols for the key delivery interface: ETSI QKD 014 and SKIP.
ETSI QKD 014 was standardized by the European Telecommunication Standardization Institute (ETSI). The other protocol is the Secure Key Integration Protocol, or SKIP, which was defined by Cisco and has recently been documented in a public IETF draft. Most QKD devices support both ETSI QKD and SKIP, and most IPsec encryptors use ETSI QKD except for Cisco encryptors that typically use SKIP. Both protocols are similar in how they operate. Note that both ETSI QKD 014 and SKIP are completely agnostic when it comes to which QKD protocol is being used. ETSI QKD 014 and SKIP are mechanisms for handing the produced key to the encryptor, regardless of the QKD protocol used to generate the key. Below is an example of a pair of QKD devices delivering a pair of symmetric encryption keys to a pair of IPsec firewalls.
This particular example is using ETSI QKD 014, but an example using SKIP would look nearly identical. Both ETSI QKD 014 and SKIP are REST interfaces that run over https. The flow of messages is as follows:
At this point, the two IPsec gateways have the same symmetric shared key and they start encrypting the traffic in the IPsec tunnel.
The image below shows how the IKEv2 protocol has been extended to transfer the public key identifier from one IPsec gateway to the other.
These extensions are standardized in RFC8784, which defines a new payload called the Post-quantum Preshared Key or PPK. The term “post-quantum preshared” might seem confusing; it's called “preshared” because RFC8784 was originally envisioned to use manually configured shared quantum-safe keys. It was only later that RFC8784 was generalized to be used for quantum keys dynamically produced by QKD devices. The protocol is simple: in the IKE_SA_INIT exchange there is a USE_PPK notification to negotiate the use of this extension. Then, in the IKE_AUTH exchange, the actual key identifier is transferred using the PPK_IDENTITY notification. RFC8784 also describes how to mix this QKD produced key with the rest of the key material, and this is shown in the box in gray on the right in the image above.
Pictured below is an end-to-end example of how all the various protocols work together to use QKD in the context of IPsec.
The sequence of events is as follows:
The IPsec gateways typically mix this QKD key with a traditional Diffie-Hellman key, and sometimes also with one or more PQC keys. A final resulting mixed key is used to encrypt the ESP data plane.
The extensions in IPsec for PQC and QKD that have been described so far are somewhat complex, but fortunately these complexities occur behind the scenes and are not visible to network operators or users. In practice, it's simple and straightforward for operators to configure and monitor PQC and QKD for their IPsec networks, as will be shown in the examples here.
Below is the screenshot of the management web interface of one widely used IPsec gateway product.
It shows the screens for configuring an IKE gateway and an IKE crypto profile. While these screens contain many configuration fields, the vast majority of these fields are things that must be configured anyway as part of the normal workflow for configuring any IPsec tunnel.
The new options for adding PQC are marked in red above. Enabling PQC for your IPsec tunnel is very straightforward and only involves a couple of new configuration fields.
The example shows the selection of Kyber-512. This is all that is required in order to turn on PQC in the network using software that is commercially available today.
Below is a screenshot of the management web interface of another widely-used IPsec product from a different vendor.
The screen for creating an IPsec tunnel pictured above shows that the only step required is selecting one of the PQC key exchange methods. In this example, ML-KEM 512 is selected. This is the only step for turning on PQC in the network using this particular vendor’s software.
Below is an example of how to configure QKD using SKIP. In this case, the example shows how to configure using QKD with the command line interface (CLI) instead of the graphical user interface.
The configuration (pictured in the background in gray above) is the configuration required for any IPsec tunnel. In black (at the front) is the additional configuration for enabling QKD. Essentially this process is as simple as indicating the need to use QKD and entering some information to identify how to reach the QKD device to get the key.
This is another example of using QKD. In this case, ETSI QKD 014 is used for the key delivery interface. Shown above is the configuration needed in addition to configuring the IPsec tunnel. A majority of this configuration is about where to get the key and how to authenticate your identity to the key delivery QKD device.
If you are using IPsec in your organization, it is important to begin planning for making your IPsec deployment quantum-safe as soon as possible. Specifically, you should learn about the technology and begin trials now to understand the impact on your organization. Securing IPsec key exchange is the most urgent migration needed to protect data-in-transit from today’s threats as well as future threats. Not only is data already at risk because of harvest now decrypt later attacks, but there are also regulatory requirements demanding these vulnerabilities be addressed in the near-term.
Deploying PQC is the baseline preparation for a post-quantum world, but if you are using IPsec to protect long-term confidential information, you should also consider deploying QKD in addition to PQC, as an additional layer of security. Applications that benefit from this additional layer of security include critical infrastructure, medical applications, financial applications, high value intellectual property protections, government data, defense security, and anything that requires long-term protection.
If you choose to deploy QKD, you have a choice of a variety of QKD protocols. Using an entanglement-based QKD protocol offers a higher level of security, better scalability, and more flexibility for deploying a general purpose quantum network that can connect a wide variety of quantum devices for applications beyond secure key distribution.
For further information, please see How to configure an IPsec tunnel using PQC keys and How to configure an IPsec tunnel using QKD keys.
As we look to the future of secure communications, entanglement-based quantum networks are at the forefront. Building these networks requires specialized software and hardware, such as beam splitters.
Entanglement-based quantum networks are being built today by a variety of organizations for a variety of use cases – benefiting organizations internally, as well as providing great value to an organization’s customers. Telecommunications companies, national research labs, and systems integrators are just a few examples of the organizations Aliro is helping to leverage the capabilities of quantum secure communications.
Aliro is uniquely positioned to help you build your quantum network. The steps you can take to ensure your organization is meeting the challenges and leveraging the benefits of the quantum revolution are part of a clear, unified solution already at work in networks like the EPB Quantum Network℠ powered by Qubitekk in Chattanooga, Tennessee.
AliroNet™, the world’s first full-stack entanglement-based network solution, consists of the software and services necessary to ensure customers will fully meet their advanced secure networking goals. Each component within AliroNet™ is built from the ground up to be compatible and optimal with entanglement-based networks of any scale and architecture. AliroNet™ is used to simulate, design, run, and manage quantum networks as well as test, verify, and optimize quantum hardware for network performance. AliroNet™ leverages the expertise of Aliro personnel in order to ensure that customers get the most value out of the software and their investment.
Depending on where customers are in their quantum networking journeys, AliroNet™ is available in three modes that create a clear path toward building full-scale entanglement-based secure networks: (1) Emulation Mode, for emulating, designing, and validating entanglement-based quantum networks, (2) Pilot Mode for implementing a small-scale entanglement-based quantum network testbed, and (3) Deployment Mode for scaling entanglement-based quantum networks and integrating end-to-end applications. AliroNet™ has been developed by a team of world-class experts.
To get started on your Quantum Networking journey, reach out to the Aliro team for additional information on how AliroNet™ can enable secure communications.