1

There obvious costs differences between them but I am not sure from a security perspective which of them considered to be more secure.

Filipon
  • 1,204
  • 10
  • 22
  • remote is usually virtualized anyway, i.e. running inside some VM on a remote host. So the question should be rather locally or remove virtualized or which underlying virtualization technique is used (i.e. different VM hypervisors, only container ...) – Steffen Ullrich Mar 18 '19 at 17:01
  • https://webgap.io/learning/virtualization-based-browser-isolation.html - You can see these guys offer some solution that is different than virtualized. – Filipon Mar 18 '19 at 17:02
  • 1
    @Filopn That's a marketing piece if I ever read one. Containerisation is basically the same thing as virtualisation, but it trades (potentially) better performance for lesser security guarantees. – Polynomial Mar 18 '19 at 17:06
  • what are those lesser security guarantees? – Filipon Mar 18 '19 at 17:07
  • I think this question is too broad. There are some vague security differences between the two overall categories, but the categories themselves are ill-defined and commonly misused by marketing, and there are further sub-categories (e.g. using hardware virtualisation vs. software, using a hypervisor vs. an OS-based option, etc.). "More secure" is also relative; secure against what criteria? Ultimately it comes down to the specifics of the solution and what your threat model is. – Polynomial Mar 18 '19 at 17:12
  • Just trying to understand the trade-offs between these two approaches from a security perspective. Obviously, these are two different approached to implement browser isolation. – Filipon Mar 18 '19 at 17:13
  • 1
    The problem is that the two approaches aren't really distinct and the security properties instead largely arise from the implementation details, which vary across solutions. – Polynomial Mar 18 '19 at 17:14
  • There might be of course gaps if you implement it in different ways but there are fundamental differences of how they operate as you see from @SteffenUllrich comment – Filipon Mar 18 '19 at 17:16
  • 1
    @Filopn: in short: no isolation is the worst. Isolation by sandboxes is better but costs more resources. Isolation by containers is even better and costs even more resources. Isolation by VM is again better and costs again more resources. And isolation by hardware is even better and costs even more resources. Of course, implementation details are relevant and might actually more security technologies to behave less secure. – Steffen Ullrich Mar 18 '19 at 17:18
  • @SteffenUllrich can you please make this comment as an answer? I am not asking about worst but trying to understand the security gaps in each. BTW, if you feel there should some edit to make sure it fits your answer please do. – Filipon Mar 18 '19 at 17:19

1 Answers1

1

In short: no isolation is the worst. Isolation by sandboxes is better but costs more resources. Isolation by containers is even better and costs even more resources. Isolation by VM is again better and costs again more resources. And isolation by hardware is even better and costs even more resources. Of course, implementation details are relevant and might actually cause more secure technologies to behave less secure.

Steffen Ullrich
  • 184,332
  • 29
  • 363
  • 424