As we have posted this Shadow Devices series, we have discussed the millions (likely billions) of new devices which will be connecting to networks over the coming decade. Clearly many of them won’t be traditional computer devices, which can be scanned and assessed for security issues. We called these other devices shadow devices because this is about more than the “Internet of Things” – any networked device which can be used to steal information – whether directly or by providing a stepping stone to targeted information – needs to be considered.
Our last post explained how peripherals, medical devices, and control systems can be attacked. We showed that although traditional malware attacks on traditional computing and mobile get most of the attention in IT security circles, these other devices shouldn’t be ignored. As with most things, it’s not a matter of if but when these lower-profile devices will be used to perpetrate a major attack.
So now what? How can you figure out your real attack surface, and then move to protect the systems and devices providing access to your critical data? It’s back to Security 101, which pretty much always starts with visibility, and then moves to control once you figure out what you have and how it is exposed.
Your first step is to shine a light into the ‘shadows’ on your network to gain a picture of all devices. You have a couple options to gain this visibility:
- Active Scanning: You can run a scan across your entire IP address space to find out what’s there. This can be a serious task for a large address space, consuming resources while you run your scanner(s). This process can only happen periodically, because it wouldn’t be wise to run a scanner continuously on internal networks. Keep in mind that some devices, especially ancient control systems, were not build with resilience in mind, so even a simple vulnerability scan can knock them over.
- Passive Monitoring: The other alternative is basically to listen for new devices by monitoring network traffic. This assumes that you have access to all traffic on all networks in your environment, and that new devices will communicate to something. Pitfalls of this approach include needing access to the entire network, and that new devices can spoof other network devices to evade detection. On the plus side, you won’t knock anything over by listening.
But we don’t see a question of either/or for gaining full visibility into all devices on the network. There is a time and place for active scanning, but care must be taken to not take brittle systems down or consume undue network resources. We have also seen many scenarios where passive monitoring is needed to find new devices quickly once they show up on the network.
Once you have full visibility, the next step is to identify devices. You can certainly look for indicators of what type of device you found during an active scan. This is harder when passively scanning, but devices can be identified by traffic patterns and other indicators within packets. A critical selection criteria for passive monitoring technology the vendor’s ability to identify the bulk of devices likely to show up on your network. Obviously in a very dynamic environment a fraction of devices cannot be identified through scanning or monitoring network traffic. But you want these devices to be a small minority, because anything you can’t identify through scanning requires manual intervention.
Once you know what kind of device you are dealing with, you need to start evaluating risk, a combination of the device’s vulnerability and exploitability. Vulnerability is a question of what could possibly happen. An attacker can do certain things with a printer which are impossible with an infusion pump, and vice-versa. So device type is key context. You also need to assess security vulnerabilities within the device. They may warrant an active scan upon identification for more granular information. As we warned above, be careful with active scanning to avoid risking device availability. You can glean some information about vulnerabilities through passive scanning, but it requires quite a bit more interpretation, and is subject to higher false positive rates.
Exploitability depends on the security controls and/or configurations already in place on the device. A warehouse picker robot may run embedded Windows XP, but if the robot also runs a whitelist malicious code cannot execute, so it might show up as vulnerable but not exploitable. The other main aspect of exploitability is attack path. If an external entity cannot access the warehouse system because it has no Internet-facing networks, even the vulnerable picker robot poses relatively little risk unless the physical location is attacked.
The final aspect of determining risk to a device is looking at what it has access to. If a device has no access to anything sensitive, then again it poses little risk. Of course that assumes your networks are sufficiently isolated. Determining risk is all about prioritization. You only have so many resources, so you need to choose what to fix wisely, and evaluating risk is the best way to allocate those scarce resources.
Once you know what’s out there in the shadows, your next step is to figure out whether and perhaps how to protect those devices. This again comes back to the risk profiles discussed above. It doesn’t make much sense to spend a lot of time and money protecting devices which don’t present much risk to the organization. But in case a device does present sufficient risk, how will you go about protecting it?
First things first: you should be making sure the device is configured in the most secure fashion. Yeah, yeah, that sounds trite and simple, but we mention it anyway because it’s shocking how many devices can be exploited due to open services that can easily be turned off. Once you have basic device hygiene taken care, here are some other ways to protect it:
- Active Controls: The first and most direct way to protect a shadow device is by implementing an active control on it. The available controls vary depending on kind of device and its embedded operating system. The most reliable means of protection is to stop execution of non-authorized programs. Yes, we are referring to whitelisting. There are commercial products available for embedded Windows devices, and a few options (some commercial, some open source) for other operating environments. There are also other defenses, including malware detection and even some forensic offerings, to really determine what’s happening on devices. Just keep in mind that active controls consume resources and can impair device stability. So you’ll want to exhaustively test any controls to be implemented on these devices to ensure you understand their impact.
- Network Isolation: These devices are on the network, which means traditional network security defenses can (and should) be used in conjunction with active controls. This involves using firewalls and/or routers & switches to provide a measure of isolation between these device networks and your data stores.
- Egress Filtering: We also recommend egress filtering on networks with these shadow devices. You can detect command and control, as well as remote access, by looking at outbound network traffic. Remember, these devices aren’t inherently malicious, so if something bad it happening it’s because an adversary is controlling the device, which means it’s communicating to a bot network or other controller outside your network, and that can be tracked.
An increasing limitation for network-oriented controls is encrypted traffic. We have published research on dealing with encrypted networks – there are ways to peek into encrypted traffic streams. When deciding between network and active controls for a devices, you need to consider encryption – network-based controls are easiest because they don’t require device interaction, but they may introduce blind spots, particularly thanks to encryption.
Finally we should mention the challenges of implementing fixes and workarounds – not only when dealing with shadow devices, but for all devices on networks now. There just isn’t enough staff to address all requirements, which means organizations need to think creatively about how to make limited staff more productive. One emerging way of magnifying staff efforts is automation. You have existing controls, which can be reconfigured based on specific rules. Of course automation of existing technologies requires integration with the existing security controls, but lately we have seen considerable innovation in this area, born out of necessity.
Our last step in figuring out your strategy to provide visibility and control over shadow devices is to determine which fixes can be automated, and establishing a strategy for that automation. We caution again that automation is a double-edged sword, which must be used carefully to ensure the ‘cure’ isn’t much worse than the symptoms. So ensure sufficient testing, and have reasonable expectations for what you can automate – in both the short and longer terms.
With that we wrap up the our series on “Shining a Light on Shadow Devices”. As we described, there will be an explosion of devices connecting to networks you are tasked to protect, and without sufficient visibility over what is connecting to them you are blind to potential attacks. There are a variety of ways to find these devices, and to implement controls protecting them. But you can’t protect what you can’t see, so make sure to shine a light on all the dark places in your networks so you are fully aware of your attack surface.