In my previous post I introduced “Goldilocks Security”, proposing three approaches to security.

Solution 1: Ignore Security

Safety in the crowd – with tens of millions of cameras out there, why would anyone pick mine? Odds are that the bad guys won’t pick yours – they will pick all of them! Automated search and penetration tools easily find millions of IP cameras. You will be lost in the crowd – the crowd of bots!

Solution 2: Secure the Cameras

For home and small business customers, a secure the camera approach simply won’t work because ease of use wins out over effective security in product design and because the camera vendors’ business model (low-cost, ease of use, and access over the Internet) all conspire against security. What’s left?

Solution 3: Isolation

If the IP cameras can’t be safely placed on the Internet, then isolate them from the Internet.

To do this, introduce an IoT Gateway between the cameras and all other systems. This IoT Gateway would have two network interfaces: one network interface dedicated to the cameras and the second network interface used to connect to the outside world. An application running on the IoT Gateway would talk to the IP cameras and then talk to the outside world (if needed). There would be no network connection between the IP cameras and anything other than the IoT Gateway application. The IoT Gateway would also be hardened and actively managed for best security.

How is this implemented?

• Put the IP cameras on a dedicated network. This should be a separate physical network. At a minimum it should be a VLAN (Virtual LAN). There will typically be a relatively small number of IP cameras in use, so a dedicated network switch, probably with PoE, is cost effective.
  • Use static IP addresses. If the IP cameras are assigned static IP addresses, there is no need to have an IP gateway or DNS server on the network segment. This further reduces the ability of the IP cameras to get out on the network. You lose the convenience of DHCP assigned address and gain significant security.
  • You can have multiple separate networks. For example, you might have one for external cameras, one for cameras in interior public spaces, one for manufacturing space and one for labs. With this configuration, someone gaining access to the exterior network would not be able to gain access to the lab cameras.
• Add an IoT Gateway – a computer with a network interface connected to the camera network. In the example above, the gateway would have four network interfaces – one for each camera network. The IoT Gateway would probably also be connected to the corporate network; this would require a fifth network interface. Note that you can have multiple IoT Gateways, such as one for each camera network, one for a building management system, one for other security systems, and one that connects an entire building or campus to the Internet.
• Use a video monitoring program such as ZoneMinder or a commercial program to receive, monitor and display the video data. Such a program can monitor multiple camera feeds, analyze the video feeds for things such as motion detection, record multiple video streams, and create events and alerts. These events and alerts can do things like trigger alarms, send emails, send text messages, or trigger other business rules. Note that the video monitoring program further isolates the cameras from the Internet – the cameras talk to the video monitoring program and the video monitoring program talks to the outside world.
• Sandbox the video monitoring program using tools like SELinux and containers. These both protect the application and protect the rest of the system from the application – even if the application is compromised, it won’t be able to attack the rest of the system.
• Remove any unneeded services from the IoT Gateway. This is a dedicated device performing a small set of tasks. There shouldn’t be any software on the system that is not needed to perform these tasks – no development tools, no extraneous programs, no unneeded services running.
• Run the video monitoring program with minimal privileges. This program should not require root level access.
• Configure strong firewall settings on the IoT Gateway. Only allow required communications. For example, only allow communications with specific IP addresses or mac addresses (the IP cameras configured into the system) over specific ports using specific protocols. You can also configure the firewall to only allow specific applications access to the network port. These settings would keep anything other than authorized cameras from accessing the gateway and keep the authorized cameras from talking to anything other than the video monitoring application. This approach also protects the cameras. Anyone attempting to attack the cameras from the Internet would need to penetrate the IoT Gateway and then change settings such as the firewall and SELinux before they could get to the cameras.
• Use strong access controls. Multi-factor authentication is a really good idea. Of course you have a separate account for each user, and assign each user the minimum privilege they need to do their job. Most of the time you don’t need to be logged in to the system – most video monitoring applications can display on the lock screen, allowing visual monitoring of the video streams without being able to change the system. For remote gateways interactive access isn’t needed at all; they simply process sensor data and send it to a remote system.
• Other systems should be able to verify the identity of the IoT Gateway. A common way to do this is to install a certificate on the gateway. Each gateway should have a unique certificate, which can be provided by systems like Linux IdM or MS Active Directory. Even greater security can be provided by placing the system identity into a hardware root of trust like a TPM (Trusted Processing Module), which prevents the identity from being copied, cloned, or spoofed.
• Encrypted communications is always a good idea for security. Encryption protects the contents of the video stream from being revealed, prevents the contents of the video stream from being modified or spoofed, and verifies the integrity of the video stream – any modifications of the encrypted traffic, either deliberate or due to network error, are detected. Further, if you configure a VPN (Virtual Private Network) between the IoT Gateway and backend systems you can force all network traffic through the VPN, thus preventing network attacks against the IoT Gateway. For security systems it is good practice to encrypt all traffic, both internal and external.
• Proactively manage the IoT Gateway. Regularly update it to get the latest security patches and bug fixes. Scan it regularly with tools like OpenSCAP to maintain secure configuration. Monitor logfiles for anomalies that might be related to security events, hardware issues, or software issues.

You can see how a properly configured IoT Gateway can allow you to use insecure IoT devices as part of a secure system. This approach isn’t perfect – the cameras should also be managed like the gateway – but it is a viable approach to building a reasonably secure and robust system out of insecure devices.

One issue is that the cameras are not protected from local attack. If WiFi is used the attacker only needs to be nearby. If Ethernet is used an attacker can add another device to the network. This is difficult as you would need to gain access to the network switch and find a live port on the proper network. Attacking the Ethernet cable leaves signs, including network glitches. Physically attacking a camera also leaves signs. All of this can be done, but is more challenging than a network based attack over the Internet and can be managed through physical security and good network monitoring. These are some of the reasons why I strongly prefer wired network connections over wireless network connections.

Next: IoT Security for Developers [Survive IoT Part 5]