How to allow only own people onto the network!

Security

✨ Staff Member ✨
Staff member
Verified Vendor
Messages
563
Likes
44
Points
980

What is NAC and how does it ​

NAC solves one problem: identifying the device connecting to the network and granting it the exact access it deserves. The system is architecturally divided into two parts. The policy server decides who the device is, whether it meets the requirements, and what permissions it has. Enforcement points implement this decision at the network level: switches, access points, and VPN gateways.

The policy server itself doesn't block anything; it merely informs network equipment which VLAN to place the port in and which access list to apply to it.

The second important mechanism is on-the-fly authorization modification. A connection check provides a snapshot of a single point in time: a device passed the check in the morning, but by lunchtime its antivirus software is disabled or its signatures are outdated. Without the ability to review access directly during a running session, such a device will remain in the working segment until the next authorization, which could mean until the end of the workday.

NAC is a policy server (RADIUS/TACACS+), with switches, access points, and VPN gateways acting as enforcement points. A device passes 802.1X using a certificate or account from the user directory, and the server issues VLANs/ACLs. Using CoA functionality, access is changed on the fly, meaning a device that violates a policy is quarantined immediately, rather than upon subsequent authorization.
Standard 802.1X authentication requires a device to have a supplicant—a software component capable of presenting a certificate or credentials. A printer, IP camera, sensor, or industrial controller typically doesn't have one, and won't. Such devices can't be left unchecked: they're often the weakest link, as their firmware is rarely updated and their passwords remain factory defaults.

A basic workaround is MAC address authentication, where the switch transmits the MAC address of the connecting device to the RADIUS server as an identifier. A more basic option is to bind an authorized MAC address directly to a switch port.

Many devices, such as printers, video cameras, and IoT sensors, do not support 802.1X. For these devices, MAC address authentication—MAC Authentication Bypass (MA)—is typically used. In this case, the switch sends the device's MAC address to the RADIUS server as a client identifier. If this MAC address is present in the RADIUS server's database, it returns authorization parameters. A simpler option is to configure port security on the switch port, which means specifying which MAC address is allowed to connect to the switch port directly in the port settings. Authentication by MAC address alone is not secure enough, as an attacker can easily spoof it.

This gap is closed by profiling—collecting indirect indicators that allow one to determine the actual device on the port. The manufacturer is determined by the MAC address prefix, the model and operating system by the contents of DHCP requests, and detailed specifications by active SNMP polling. The key here is consistency: if the profile is checked constantly, and not just at the moment of connection, substituting a laptop for a printer becomes significantly more difficult.

A separate category is represented by employee personal devices and contractor equipment. The problem here is not the lack of a supplicant, but the lack of control over the device itself: there is no guarantee that corporate security tools are installed on a personal laptop, and its status cannot be checked by an agent. The solution is based on isolation—such connections are placed in a separate segment with a limited set of access rights, and expanding rights requires a registration procedure.

Personal, guest, and unregistered devices are isolated in separate segments with limited access. Self-registration portals are used to obtain extended privileges. Since it is not always possible to install a NAC agent on such devices and verify the presence of corporate information security tools, special care is required when configuring access policies.
The first thing that eliminates options is compatibility with the equipment already in the network. NAC operates through switches and access points, and if part of the fleet doesn't support the required mechanisms, the system simply won't be able to implement anything in those areas. Therefore, the assessment begins with an inventory list, rather than a comparison of product features: which of the existing ones supports 802.1X, which supports MAC authentication.

Next come characteristics that manifest themselves at scale: operation with distributed sites under unified policies, resilience under peak loads when everyone connects simultaneously on Monday mornings, and the heterogeneity of the endpoint fleet.

The choice of solution is always individual and should be based on business objectives and infrastructure specifics. It is important to consider the support of the required network equipment control mechanisms, the diversity and size of the endpoint fleet, the connection types used to cover all access scenarios, the ability to centralize management and enforce unified access policies in geo-distributed environments, and the guarantee of stable NAC operation under peak loads.

The third set of criteria is integration. NAC makes decisions based on the user's identity and group, so integration with the user directory and access management system is essential. Connection and quarantine events must be reported to the SIEM, otherwise control remains locked within a single system. Another feature that's often overlooked when choosing a solution is administrative access control to network equipment: this falls into the same category of solutions.

Criteria: support for a real hardware fleet, scalability to distributed sites, integration with user directories, IdM, and SIEM, agent-based endpoint health checking with automatic quarantine. Administrative access control via the TACACS+ protocol is also worth noting; this is also part of NAC functionality.
For organizations working with government information systems, critical infrastructure facilities, personal data, or industrial control systems, NAC addresses several mandatory measures at once: identification and authentication, access management, and security event logging.

This means that the solution must meet not only technical but also formal requirements: FSTEC certification and inclusion in the registry of domestic software.

The second requirement is not related to the regulator, but to the state of the equipment itself. For most companies, this state is mixed: some switches and access points have already been replaced with Russian-made ones, while others remain imported and will continue to function for many years to come. A solution that supports only one of these components leaves part of the network uncovered and effectively imposes a hardware replacement schedule dictated by product limitations rather than the company's plan.

For GIS, critical information infrastructure, personal data, and automated process control systems, NAC covers identification, access management, and event registration measures under FSTEC orders No. 117, 21, 31, and 239. Therefore, it's important to ensure that the solution is FSTEC-certified and included in the Ministry of Digital Development's registry. Furthermore, to ensure that the solution itself doesn't become an obstacle to import substitution, it must support both Russian and imported equipment (if you still have any) and operate on a Russian technology stack.
The project begins with an inventory, and this isn't a formal step. It's the collection of initial data, without which policies cannot be written. It's essential to understand which equipment supports which control mechanisms, where users connect, and what network accesses they actually use. Next comes design: an access matrix, policies, and configuring integrations with the user directory, access management system, and SIEM. Only then does deployment begin, segmented, with testing and debugging at each stage.

Roles here are divided between information security, which leads the project and creates policies, and IT, which is responsible for the network and equipment configuration. A third essential participant is service owners and department heads: they confirm which accesses for their employees are considered legitimate; without this, the access matrix is a matter of chance.

The project begins with an inventory and infrastructure survey: which equipment supports 802.1X, MAB, and CoA, where users connect, and what network access points they use. Based on this data, a design is then developed, an access matrix is created, and policies are written. At this stage, integration with other systems—user directories, SIEM, and IdM—is also configured. Only then does the product's segmented implementation begin, with testing and debugging of the policies developed in the previous stage.

The key to implementing the system is to avoid immediate enforcement. The system initially operates in observation mode, collecting data only: which devices are online, which are noncompliant, and who is using which devices. This picture almost always diverges from expectations, and it becomes the basis for realistic policies.

Then, the expansion proceeds gradually: initially, a limited number of departments or sites, with lenient rules, and only as data accumulates are the rules tightened. This pace may seem slow, but it pays off in the absence of widespread access disruptions.

Integrating NAC into your infrastructure requires a well-thought-out strategy and gradual scaling. It's best to start with a read-only mode for collecting device data to identify policy inconsistencies and understand which users and devices are accessing which network areas. It's recommended to initially apply policies to a limited number of departments, equipment groups, or locations, using a lenient set of rules and based on the collected data. Requirements should be tightened and policies adjusted gradually.
The main mistake is trying to roll out NAC to the entire infrastructure at once, without a pilot or policy debugging. The day after deployment, access begins to be denied to those not considered during the design, the number of support requests grows, and under the pressure of discontent, the rules are relaxed.

The main mistake when implementing NAC is trying to do too much and force the entire infrastructure into the new solution without proper design and testing. This approach, at best, results in having to start over, gradually and consistently, after a series of errors and user complaints. At worst, errors are resolved with "allow all" policies, which defeat the purpose of using a NAC solution.

The impact of NAC is measured by direct metrics and becomes visible even before full deployment—in monitoring mode.

The first metric is the share of devices not in the inventory: the system displays the actual number of connections, and the gap between it and the inventory is usually significant. Reducing this gap is the first measurable result.

The second is the share of devices that do not meet security requirements: without an agent installed, with disabled protection, or with outdated updates. The dynamics are important here, as they indicate whether policies are working or whether they are being circumvented.

The impact of NAC can be assessed during implementation based on the dynamics of several metrics, including the share of unregistered and non-compliant devices and a reduction in the number of incidents, including those related to unauthorized access.
NAC answers a question that, before its introduction, simply wasn't asked in most networks: what exactly is currently connected and why. The system is architecturally simple—the policy server makes the decision, and network equipment executes it. But the outcome depends on how precisely the policies themselves are defined, and they're impossible to write without a complete picture of who's using the network and where.

The complex part of the project isn't the corporate laptops, but everything else. Printers, cameras, sensors, and industrial controllers can't authenticate, so they have to rely on a combination of MAC address authentication and profiling, where the former is easily spoofed, and the latter only classifies the device but doesn't confirm its authenticity. Personal and guest devices are addressed differently—not through control, but through isolation and access restriction.

When selecting a solution, the verification process is the reverse of the usual: first, compatibility with existing network equipment, then functionality. A feature-rich product is useless in areas where switches don't support the necessary mechanisms. For regulated industries, this is supplemented by the formal component—FSTEC certification and a registry—as well as the ability to work simultaneously with both Russian and remaining imported equipment.

Results should be assessed based on dynamics, not the actual implementation. A reduction in the share of unregistered devices, a decrease in the number of non-compliant connections, and a reduction in unauthorized access incidents are measurable indicators that become apparent during deployment. If these figures remain stagnant six months after launch, the issue isn't the product, but rather the extent to which politicians have become silly.