Re: Canal+ crash

[Niklas Holsti <niklas.holsti@tidorum.invalid>]

On 2024-07-21 12:19, Dmitry A. Kazakov wrote:
> On 2024-07-21 10:00, Niklas Holsti wrote:
>> On 2024-07-21 10:22, Dmitry A. Kazakov wrote:
>>> On 2024-07-21 03:04, Lawrence D'Oliveiro wrote:
>>>> On Sat, 20 Jul 2024 11:08:47 +0200, Dmitry A. Kazakov wrote:
>>>>
>>>>> On 2024-07-20 09:43, Lawrence D'Oliveiro wrote:
>>>>>
>>>>>> On Sat, 20 Jul 2024 09:23:11 +0200, Dmitry A. Kazakov wrote:
>>>>>>
>>>>>>> It is about the fundamental principle that security cannot be 
>>>>>>> added on
>>>>>>> top of an insecure system.
>>>>>>
>>>>>> Actually, it can. Notice how the Internet itself is horribly 
>>>>>> insecure,
>>>>>> yet we are capable of running secure applications and protocols on 
>>>>>> top
>>>>>> of it.
>>>>>
>>>>> Why on earth do we need security updates?
>>>>
>>>> Because computer systems are complex, and new bugs keep being 
>>>> discovered
>>>> all the time.
>>>
>>> This does not make sense. You can create a very complex system out of 
>>> screwdrivers and still each screwdriver would require no update.
>>>
>>> Systems consist of computers and computers of software modules. There 
>>> is nothing inherently complex about making a module safe and bug 
>>> free. Security interactions are primitive and 100% functional. There 
>>> is no difficult issues with non-functional stuff like real-time 
>>> problems.
>>
>> Well, several recent attacks use variations in execution timing as a 
>> side-channel to exfiltrate secrets such as crypto keys. The crypto 
>> code can be functionally perfect and bug-free, but it may still be 
>> open to attack by such methods.
> 
> It is always a tradeoff between the value of the information and costs 
> of breaking the protection. I doubt that timing attack are much more 
> feasible in that respect than brute force.


Security researchers and crypto implementers seem to take timing attacks 
quite seriously, putting a lot of effort into making the crucial crypto 
steps run in constant time.


>> But certainly, most attacks on SW have used functional bugs such as 
>> buffer overflows.
> 
> Exactly. Non-functional attacks are hypothetical at best. They rely on 
> internal knowledge which is another problem. 


As I understand it, the "internal knowledge" needed for timing attacks 
is mostly what is easily discoverable from the open source-code of the 
SW that is attacked.