Kindly have a look at the following survey articles:

Article A survey on gaps, threat remediation challenges and some tho...

Article An Intrusion Detection And Prevention System In Cloud Comput...

Article Mobile Edge Computing, Fog et al.: A Survey and Analysis of ...

I hope these articles will help you.

Thanks

I think a practical step is to obtain data as much as possible. If there are enough data, you can easily continue your research on various models and algorithms.

Chukwudi,

Before answering those questions you have to answer the following:

1. What do I want to focus on this broad question? Some of the parameters to contemplate:

• Type of cloud environments- infrastructure as a service, platform as a service, application as a service
• ML techniques- supervised unsupervised, reinforcement.
• Type of attackers - internal, external, consultant
• Particular types of vulnerabilities and exploits
• Type of assets- money, infrastructure for further attacks, etc.

2. Is my focus going to be on machine learning or security? If the focus is on machine learning with security applications then your most likely going to focus on increasing accuracy on cloud based scenarios. If you choose security as primary topic then it is a question of what cloud scenarios are available in which machine learning could prove useful. these are two very different approaches.

3. Does the data need to be real or synthetic? The data does not have to be necessarily gathered from production environments. The data might come from honeypots(in which case bootstrapping may help. Also other setups might focus on particular threat models and from there synthetic data may be generated based on the threat model

The answer to these questions will provide the answer that you are looking for.

hope this helps