Construction of a CPS requires a multi-dimensional team to ensure that it is constructed safely and securely. The authors Wolf and Serpanos  suggest a number of solutions for design and run-time analysis of the CPS, pulling from both hardware and software test and evaluation realms. The techniques seem reasonable for larger or more costly CPS systems but would be hard to implement across a CPS built from IoT devices.
The current expectation for IoT devices is that they have limited power and computing resources so as to continue to drive down cost, making Massive IoT possible. Limited computing resources means that these devices will have been designed with trade offs between the capabilities offered by the device and the ability of the device to function with safety and security in mind. In other words, these devices will be the weakest link in the CPS.
Massive IoT deployments are separate from Critical IoT deployments in terms of the cost and performance. Devices targeted for Massive IoT deployment typically have less computing resources whereas Critical IoT devices have more resources to guarantee the timeliness, reliability and safety of the data transmitted. Data for both of these systems is transmitted over some sort of network medium to a control or consumer station. The validity of the data coming out will only be as good as it needs to be to meet market demand. In other words, transmission of my footsteps from my Fit Bit is not as important as haptic feedback for remote surgery.
Massive IoT devices are going to be insecure by their market drivers and efforts to change the market influences is outside the scope of this paper. Critical IoT devices are going to do more for security and safety but be at odds with the latency requirements. Therefore, these devices will provide different levels of protections inherently.
The data from these devices flow through one or more transport mediums, each with their own weaknesses due to their market drivers. At the end of the day, data received at a consumer level may or may not be the data that was actually generated at the sensor level. The amount of time and cost put in to determining plausibility or increasing veracity of the sensors is dependent upon costs determined by an appropriate risk model. In the end, the data is only as valid as you need to to be.
The CPS or ICS developers can use a holistic risk assessment when looking to implement security and safety during design time and run-time. Design time implementations increase deployment costs for IoT devices while run-time implementations can be deployed over time and lowering the initial deployment cost.
It would seem that IoT manufacturers may want to develop hardened and consumer level products that would give flexibility to the deployers. The hardened model will offer safety and security features in the device itself whereas the one that does not will be left up to the end user. The benefit of this is that it provides different cost models to individuals (persons or corporations) that wish to deploy the systems. Bad news is that the weakest link then becomes the least secure device on the network. Perhaps carriers (ISP, eg) then charge more for the risk associated with transmitting the information. This leads to economic challenges as security and safety could be viewed as pay-to-play in this scenario.
A CPS comprised of IoT devices that prioritizes safety and security of the overall system will be more complex than one that does not. IoT devices in and of themselves will be driven by market demand to fulfill the need for Massive IoT deployments. The more security and safe the system is to be, the higher the cost as additional features and functionality are pushed into the end device. Various techniques can be used to increase safety and security from the end device all the way to the control station and back but come at a cost that is evaluated against an acceptable level or risk for the application itself.
 Wolf, Marilyn, and Dimitrios Serpanos. “Safety and Security in Cyber-Physical Systems and Internet-of-Things Systems.” IEEE, January 2018.
 Akpakwu, Godfrey Anuga, Bruno J. Silva, Gerhard P. Hancke, and Adnan M. Abu-Mahfouz. “A Survey on 5G Networks for the Internet of Things: Communication Technologies and Challenges.” IEEE Access 6 (2018): 3619–47. https://doi.org/10.1109/ACCESS.2017.2779844.
 Shariatmadari, Hamidreza, Rapeepat Ratasuk, Sassan Iraji, Andrés Laya, Tarik Taleb, Riku Jäntti, and Amitava Ghosh. “Machine-Type Communications: Current Status and Future Perspectives toward 5G Systems.” IEEE Communications Magazine 53, no. 9 (September 2015): 10–17. https://doi.org/10.1109/MCOM.2015.7263367.