2

I am writing my bachelor thesis in CS on cloud security and its weaknesses and started to do research today and got quite overwhelmed.

I am mainly looking at IaaS and PaaS services and have decided to focus mostly on threaths caused by attackers. But it is really hard to decide where to start.

What is the biggest security issues with cloud atm? Would be nice to hear from someone who has worked more with cloud than I have. Is there any scientific article released in recent years discussing this? I only managed to find older ones.

Should I focus on encryption/decryption and what can be done there or are there bigger more pressing issues in cloud security? Any advice woul help.

Some articles I found interesting so far: https://downloads.cloudsecurityalliance.org/initiatives/top_threats/The_Notorious_Nine_Cloud_Computing_Top_Threats_in_2013.pdf

http://www.enisa.europa.eu/activities/risk-management/files/deliverables/cloud-computing-risk-assessment

But even here there is just too much fo someone rather unfimiliar with the subject to write WELL about in 4 weeks. So can anyone with experience recommend the more common security risks that could be good for me to focus on.

Green_qaue
  • 155
  • 5
  • The point of a scholarly research is to sift through the mess to focus on what is relevant. This is your first day, be patient. Keep reading. Your focus on IaaS/PaaS is STILL too broad. Keep reading and you will find an avenue to explore that will help you focus. – schroeder Apr 25 '14 at 19:02
  • You need to devise a threat model before you can evaluate the risks posed. I'd suggest that most threat models for "cloud security" should include distrusting the service provider--and, I suspect, therein will lie the most realistic and most serious threats. – eggyal Apr 25 '14 at 19:10

1 Answers1

2

From the cryptographic side of things, one big topic these days is side channel attacks: attackers try to extract information on private values in a given system by observing side effects such as power consumption, precise timing... A synthetic view of such attacks is that there are two distinct "worlds":

  • the "abstract world" in which a system receives inputs as sequences of 0 and 1, and outputs similar sequences;
  • the "physical world" in which computation takes place within measurable constraints such as computation time.

Side-channel attacks are about exploiting the differences between these two worlds.

The cloud is an attempt at disbelieving this difference. With the cloud, you get virtual machines under the assumption that they are indistinguishable as physical ones, since they compute the same things. Indeed, a VM will respond with the same sequences of 0 and 1 for the same inputs, so in the abstract world, a VM is a perfect emulation of a physical machine. But if we take notice of physical details, some problems arise. In particular the two following:

  • For many cryptographic protocols, there is a need for randomness. Randomness is extracted from physical elements. With a VM, the "physical" elements are emulated. A consequence is that a PRNG within a VM may be of suboptimal quality.

    Typically, consider what happens if you take a snapshot of a live VM, and (later on) restore that snapshot: it is quite possible that the VM's PRNG will output the exact same sequence of random values that it did the first time, which can lead to severe security issues. An extreme example is the generation of the "k" value in DSA and ECDSA signatures: reusing the same k for two signatures (on distinct messages) reveals the private key (see this answer).

  • A VM run on some hardware, and that hardware may host several VM simultaneously. These VM share the same resources, in particular caches. This allows for cross-VM cache timing attacks.

    This has been recently demonstrated in lab conditions: the researchers managed to pull off a cache timing attack on some AES implementations, allowing a VM to guess an AES encryption key used in another VM which just happened to run on the same hardware.

Side-channel attacks on cryptographic algorithms do not exhaust the vast well of security issues related to cloud computing, but they are still a fertile investigation field, and are enough to warrant a cautionary warning: you really really do not want to see your VM run on the same hardware than the VM from your competitors.

Thomas Pornin
  • 320,799
  • 57
  • 780
  • 949
  • Great answer, thank you :) Will take a look at the links. "you really really do not want to see your VM run on the same hardware than the VM from your competitors." Is that because the risk of the competitors using this cross-VM cache timing attacks? – Green_qaue Apr 25 '14 at 19:12
  • Yes, that's what I say. Also, a VM can be disruptive by, for instance, saturating the common memory bus. You really don't want your neighbouring VM to be evil and hostile. – Thomas Pornin Apr 25 '14 at 19:16