Fellow/.ers may recollect the
"cool scientific paper" where the authors bomb DRAM chips with Xrays or more simply with 50-watt spotlight bulb to exploit Java and.NET virtual machines. That attack is very relevant to this new scheme proposed in this RFC.
What if the new security bit flips? Wow! I just improved my result from 70% to 100% !!! I should waste no time in typing my latest paper.
I agree with your sentiments. 1)If typechecking is done at every reference, then for each field in the program, the machine shud keep track of its type. So if i have 100M fields(pointers) in my machine, then it means that I need to keep track of 100M types. And this is way too much overhead.
2) It does not matter whether or not the implementation is entirely conrained inside a VM. all i need to do is to turn off the security mangaer, which can be done from inside the virtual machine.
3) I agree that compiler optimisations help my attack.
btw, i dunno if you checked my comments on the comments posted here. you can get my comments here
Do email me about the results of your experiments.
--Sudhakar.
Yes you are rite, but not rite.
The program makes no use of types at runtime. I mean, there are no typecasts in the program. Hence, the typesafety of the applet can be verified at link time. My guess is that if type safety can be verified at link time, a virtual-machine will try to do it at link time. Finally, a lot depends on the exact implementation details of the VM. I have not looked at Smalltalk and Lisp. So I cant say much about them. However, my instinct tells me that some attack can be launched in those cases. Please let me know if you have some ideas! I am still undecided about what happens with dynamic checking. All depends on what exactly I mean by dynamic checking.
Sudhakar.
It was a pleasant surprise to see my paper on/. this morning. Now pdf slides are available here . My comments on the views shared here are also available .
Sudhakar .
Slashdot Mirror
its boring to read the same article the 4th time!
moderators, wake up !
Fellow /.ers may recollect the
"cool scientific paper" where the authors bomb DRAM chips with Xrays or more simply with 50-watt spotlight bulb to exploit Java and .NET virtual machines. That attack is very relevant to this new scheme proposed in this RFC.
What if the new security bit flips? Wow! I just improved my result from 70% to 100% !!! I should waste no time in typing my latest paper.
--Sudhakar.
I was a bit lazy with keys, to save the pressure on my fingers. Anyway, I see what you are saying.
--Sudhakar
I agree with your sentiments. 1)If typechecking is done at every reference, then for each field in the program, the machine shud keep track of its type. So if i have 100M fields(pointers) in my machine, then it means that I need to keep track of 100M types. And this is way too much overhead. 2) It does not matter whether or not the implementation is entirely conrained inside a VM. all i need to do is to turn off the security mangaer, which can be done from inside the virtual machine. 3) I agree that compiler optimisations help my attack. btw, i dunno if you checked my comments on the comments posted here. you can get my comments here Do email me about the results of your experiments. --Sudhakar.
Finally, yes I bothered to read the paper. I might have read it some 100 times. After all I wrote it. :-)
Sudhakar.
Yes you are rite, but not rite. The program makes no use of types at runtime. I mean, there are no typecasts in the program. Hence, the typesafety of the applet can be verified at link time. My guess is that if type safety can be verified at link time, a virtual-machine will try to do it at link time. Finally, a lot depends on the exact implementation details of the VM. I have not looked at Smalltalk and Lisp. So I cant say much about them. However, my instinct tells me that some attack can be launched in those cases. Please let me know if you have some ideas! I am still undecided about what happens with dynamic checking. All depends on what exactly I mean by dynamic checking. Sudhakar.
It was a pleasant surprise to see my paper on /. this morning. Now pdf slides are available here . My comments on the views shared here are also available .
Sudhakar .