Re:Don't run 'strings' on untrusted files
by Animats on Saturday October 25, 2014 @03:00PM
(#8509012)
Address space randomization is security theater. It slows down attackers, but doesn't stop them. See "stack spraying". What it does prevent is replication of security bugs, which allows vendors to ignore them.
Re:Don't run 'strings' on untrusted files
by vezzy-fnord on Saturday October 25, 2014 @03:16PM
(#8509058)
If it slows them down and makes their lives harder, then it is a worthy mitigation. That's what a mitigation is, after all: it reduces severity and effectiveness. It's not a full stopgap. Even still, ASLR implementation quality has left much to be desired throughout the years.
Re:Don't run 'strings' on untrusted files
by wyager on Saturday October 25, 2014 @03:06PM
(#8509023)
Counter-ASLR measures are not always viable. Regardless, Go also lacks DEP, which is the more important of the two.
Re:Don't run 'strings' on untrusted files
by imanaccount247 on Saturday October 25, 2014 @03:22PM
(#8509081)
No it isn't. If trying to exploit some daemon requires thousands of tries instead of one try, I am thousands of times more secure. Because they will almost certainly fail the first time, causing the daemon to crash, and then have no more tries. It is only security theater if you live in the land of "lets use broken software that crashes all the time and then run a monitoring tool to automatically restart it". That is the problem.
Or, more generally, memory safety.
Of the three you described, Rust makes the strongest memory safety guarantees. Neither D nor Rust require a complicated runtime. I'd say this isn't really Go's intended area of usage.
Also, Go is missing things like ASLR and DEP (last time I checked), which means that if you link to any vulnerable non-Go code using cgo (which is almost inevitable when writing core utils), or if you find a good bug in the Go runtime, it's trivial to get code execution.