17 reasons not to trust IoT … yet

Just as drivers trust their cars to respond to turns of the steering wheel or taps on the brakes, the IT community must be able to trust the internet of things. With the ever-increasing number of connected devices that feed data to monitoring systems, modern society increasingly depends on the IoT.

To help define trust within IoT systems, the National Institute of Standards and Technology has issued a draft report listing "technical concerns" that can undermine trust in IoT products and services.

Although NIST acknowledges trust in the data a connected device receives is important, this report, primarily addresses the level of confidence humans can have that IoT services perform as intended. The 17 concerns – and some possible remedies – are as follows:

1. Overwhelming scalability. It's not the billions of "things" that will comprise IoT networks, it's the complexity of the connected communications because it becomes impractical to test with any degree of thoroughness. By considering sub-networks or setting some boundaries on IoT systems – such as limiting their access to the internet – the threat surface can be reduced and trust can be established for specific components rather than for the entire IoT network.

2. Heterogeneity. Competition in the IoT marketplace will lead to incompatible systems and "things" that cannot be easily replaced in a closed architecture, creating unforeseen security vulnerabilities. Additionally, as heterogeneity increases, so will risks to the supply chain as it becomes more difficult to verify the authenticity of a component. 

3. Loss of control and ownership. When much of the functionality within an IoT system originates from third-party vendors, IoT network owners have little ability to verify the integrity of "black box" components.

4. Composability, interoperability, integration and compatibility. Unlike military or safety networks, IoT systems may be built without a full understanding the actual security of the "things" and their environment and communications channels.

5. Heterogeneity. When it comes to IoT systems, there are dozens of "ilities" -- availability, compatibility, discoverability, durability, flexibility, etc. -- that are not easily measured. It's difficult to know which “ilities” are more important (security or performance?), and cost is not well understood. To ensure an IoT network's intended behaviors match actual performance, developers should consider the  “ilities” at the beginning of the IoT life-cycle.

6. Synchronization. Like air-traffic control, the operation of IoT networks depends on the synchronization of calculations, data transfers and events running across numerous distributed computing systems. No practical, trusted universal time-stamping mechanism currently exists for IoT networks, so timing anomalies will occur, enabling vulnerabilities, poor performance and network failures.

7. Lack of measurement. Because IoT comprises a relatively new set of technologies, few metrics and measures are available, making it difficult to argue that a system is trustable or even estimate the amount of testing that a system should receive.

8. Predictability. At a fundamental level, useful IT systems depend on predictability. Because an IoT node's location, signal strength and data transmissions – among other characteristics -- can vary, it's difficult to predict that required components, resources and functions will be available when needed.

9. Few IoT-specific testing and assurance approaches. The huge numbers of interdependencies and "black box" components in IoT systems make them difficult to test, as do the multiple potential sources of failure. A 99.9 percent reliability rate in one component might be acceptable, but not if it is one of a dozen "things" with the same reliability in a critical safety or medical network. Although redundancy will reduce risk, it also adds complexity.

10. Lack of IoT certification criteria. Certification of products is typically complex, expensive and time consuming, especially when threats and operational environments are not completely known. Additionally, even if a "thing" is certified, its use in an IoT system where not all components received the same type of vetting can undermine functionality.

11. Security. Besides the security of the individual components, IoT networks are also subject to vulnerabilities related to default credentials and difficulties with patching and upgrading. However, solutions that explicitly restrict IoT communications to sources and destinations intended by the manufacturer along with the use of transport standards for secure upgrades can boost the overall security of devices.

12. Reliability. The changing nature of IoT networks and the complexity of the relationships among the components create new risks and vulnerabilities, making make it nearly impossible to determine a system is reliable. Determining the cause and liability for system failure is likewise problematic.

13. Data integrity. The accuracy, fidelity and availability of data along with its timeliness and hosting location directly impacts whether an IoT system is trustworthy. Specific data integrity concerns include missing or incomplete data, data quality, faulty interfaces, data tampering, security and privacy and data leakage.

14. Excessive data. Because IoT systems will generate massive amounts of data, it will be extremely difficult to isolate and treat corrupt data in IoT systems quickly enough.

15. Speed and performance. As data generation and computation speeds up, real-time forensics and recovery from faults and failures will become more difficult, making it harder to correct errors and data anomalies. Additionally, there are currently no simple speed metrics for IoT systems.

16. Usability. Engineering tradeoffs that make IoT devices easy to use inherently affect security. Many smart devices have limited or no direct user interface, so changing passwords or updating software is problematic. Consistent user interfaces would help overcome challenges to providing effective security.

17. Visibility and discovery. As IoT devices become ubiquitous, the technology that runs them tends to disappear from view, making it difficult to trust, for example, voice-activated appliances that must always be listening for commands. Further, the IoT has no standard communications protocol, which has led to a variety of proprietary technologies, which adds further complexity and limits device discovery and overall trust.

NIST is looking for feedback on the 17 technical concerns as well as suggestions for other issues that may be missing from the document.  Comments will be open until Nov. 5.