October 14, 2020
We expect media access control (MAC) address randomization to be the default setting on more than 30% of mobile devices by Feb 2021. Since many network service providers (NSP) use MAC addresses as sole unique device identifiers, randomization will impact quite a few of their systems.
No Real Way to Identify Devices
MAC randomization will primarily impact device intelligence – the NSP’s ability to identify device types and models – a key element in most management, troubleshooting, and diagnostics solutions. Customer experience might also take a hit, leading to more customer service calls and greater demand for hardware, such as mesh extenders.
Traditionally, these solutions relied on the easiest way to identify devices – MAC addresses – but now, as iOS 14 and many Android 11 devices come with MAC randomization ON by default, many network service providers might struggle with any of the following issues.
Harder to Provide Good Home Network Speeds
Slow network speeds are a major issue for network service providers: it is the sole reason behind a significant number of customer service calls. After all, a good internet service provider (ISP) is one you can pay for and forget about.
A good internet service provider is one you can pay for and forget about.
To give customers a smoother and faster home network experience, ISPs turned to mesh networks for ‘whole home’ coverage. A mesh network is a set of Wi-Fi extenders that need precise device analytics to work as efficiently as possible with the smallest number of extenders for cost efficiency.
Mesh networks not only provide good coverage and speed, leading to more satisfied customers and fewer service calls, they also provide an opportunity to sell an additional service. Nevertheless, managing a Wi-Fi mesh network requires good analytics and device statistics, and these solutions usually rely on MAC addresses.
Mesh Network Troubleshooting Relies on MAC Addresses
Troubleshooting a mesh network works best with precise device identification, normally based on MAC addresses. It detects patterns in device connectivity over time. For example, a mobile device might have a bad connection in a single room only, thus moving an extender would make the network perform better without any additional hardware. If the device randomizes the address, it becomes harder to pin down the issue.
Also, devices that randomize MAC addresses mask their device type. This is not an issue for machine-learning powered solutions such as CUJO AI Explorer, but MAC-based device identification will not be reliable for Wi-Fi speed and connectivity statistics. Case in point: if an extremely old Android device is getting poor speeds due to its hardware limitations, the network has nothing to do with it, but the NSP cannot determine that the device is to blame without a way to identify and classify it.
Proper device intelligence would show the network service provider that sending the client an additional mesh extender would not do anything good for anyone: neither the speed of the device, nor the company’s bottom line.
Fewer Patterns for Finding Dark Spots in Home Networks
Mesh extenders are expensive, but the largest financial risk stemming from MAC randomization is the guarantee NSPs provide to their customers. Some leading NSPs guarantee to provide 100% connectivity on home Wi-Fi and charge additional service fees.
Network analytics based on individual devices can show patterns in connectivity and help network managers discover dark spots – areas where connectivity is worse. If your phone gets a weak Wi-Fi signal every time you are at the garage, you can modify your networks layout to cover any weaker area. These patterns show up in network analytics tools thanks to identifiable devices: the analysts can figure out that it’s precisely your device that struggles with getting a proper connection at a certain time.
On the other hand, if your network manager cannot determine whether it’s only your phone connecting poorly at that time, the network will seem like it needs a complete overhaul or more hardware than it does. This happens when a single device randomizes its MAC address and shows up as separate devices in the network log.
Nevertheless, troubleshooting a network is a reactive stance, and many leading NSPs are looking to preempt any customer dissatisfaction with the help of proactive diagnostics. As you might’ve guessed, these are also impacted by MAC randomization.
Poorer Proactive Diagnostics
Reacting to connectivity issues causes frustration for customers as well as NSPs, who are forced to manage larger numbers of customer calls and complaints. Proactive diagnostics prevent customer dissatisfaction and reduce the number of calls an NSP gets. Their sole goal is to predict and prevent connectivity issues that the customer might not even be aware of before they becomes pressing.
Large network service providers are investing significant amounts of money and resources into machine learning solutions that analyze network data and look for complex patterns. These algorithms depend on precise data points and statistics, such as device types, models, and operating systems that are active on their Wi-Fi networks.
As the networks rely on MAC addresses to identify devices, randomization might severely impact their predictive proactive machine learning models, making them unreliable. On the other hand, using machine learning solutions is exactly how we’ve managed to solve device identification on the model and OS level at CUJO AI without relying on MAC addresses.
Unreliable Home Network Management Apps, Parental Controls, and Pause Internet
Many internet service providers offer apps to help home users manage their network. Some apps allow users to track internet usage on the device level or control access to the internet for children. Pause internet – a timeout for connections – is also in demand for families that crave an easy solution for screen-less time on a schedule.
As you might have already guessed from the context of this article, these solutions also need a way to identify individual devices. In most cases, this means relying on MAC addresses. Thanks to randomization, NSPs will struggle to provide reliable data in their apps. It is also likely that many clients will stop using these apps due to low data accuracy and poor usability.
Poor usability might become a predominant issue for parental controls – if a minor’s device randomizes its MAC address, the network will no longer prevent it from accessing mature content.
An Even Worse Customer Experience with Captive Portals?
Captive portals are most often used in public Wi-Fi networks, but some NSPs use them for new user sign-ups, billing or user activation flows. Since these solutions identify devices by their MAC addresses, captive portals could start popping up for users that have already agreed to the terms of service or passed their activation procedures. For some users, randomized MAC addresses might force captive portals to completely break their online experience.
Moving Forward without MAC Addresses
CUJO AI has recently patented the new Device Intelligence solution that can identify over 50,000 unique mobile and IoT devices without relying on MAC addresses. Instead, our algorithms analyze over a dozen metadata points to identify device types, manufacturers, OS and firmware versions with over 74% accuracy in the first 5 minutes. After the first 24 hours, this percentage increases to over 92%.
Read more about how we do it in our next blog post about CUJO AI solving device intelligence without MAC addresses.
Alternatively, download our whitepaper on MAC randomization.