1) The principle issue in application layer related to iot security is DDOS attack
This paper gives more details about this attacks and how to detect it:
Zolotukhin, M., Hamalainen, T., Kokkonen, T., & Siltanen, J. (2016). Increasing web service availability by detecting application-layer DDoS attacks in encrypted traffic. 2016 23rd International Conference on Telecommunications (ICT). doi:10.1109/ict.2016.7500408
2) I think blockchain technology is the best solution for critical iot security problems
'IoT security: Review, blockchain solutions, and open challenges' , Minhaj Ahmad Khan, Khaled Salah
IoT is by nature starting from B2B, and from M2M. Humans come afterwards.
Therefore the authentication is first not related to humans, but to processes, to sub-systems. Robustness is essential because with IoT, everything starts in the physical world, goes to the digital representation space for efficient and automated processing, and feeds back to the real physical world, which it acts upon.
Therefore an end to end approach of what the system does is essential to any consideration of security.
That helps define Threat scenarios, and Adversarial Models, then comes the possibility to work on response strategies. These can combine methods such as the three mentioned in the question by Jaydeep Gheewala .
In my opinion, it depends on the use case of IoT, if we want IoT nodes with long life time , we should reduce electrical energy consumption, then the node processor and memory should be low performances, however block chain and certificate based approachs will need high computing and memory performances.
To add security in IoT kind application many options are available if you are interested in cryptography then go for lightweight techniques to ensure it as power is an important resource for the network lifetime.
I suggest certificates are a very good approach for this but it has management overhead for good understanding you can refer my paper
Title: Incorporating privacy and security in military applications based on opportunistic sensor networks and feedback be if you find it useful. If you find an alternative you can discuss it with me.