Talk:Security

IMHO, the endusers should have to spend zero time worrying about security. The laptops have to be as trustworthy as a book.

--BobBagwill 09:20, 13 April 2006 (EDT)

Just as you have to keep a book in your physical possession to guarantee it does not get written on, soiled, or destroyed, physical possession of the laptop will always allow malicious individuals to defeat its security. So trustworthy as a book and secure as a book are excellent goal statements but they do not equate to zero time worrying about security.

--The Guy who used Negroponte's Q&A session to push Domestic Adoption

I disagree. If the OS is digitally signed, you can guarantee it hasn't been tampered with. If the laptop will only download/store/boot/execute signed images, updates will be trustworthy too. That leaves user data. Assuming the only authentication will be a password or challenge/response, the laptop will be vulnerable to shoulder-surfing. If the laptops backed up to a central server, you could always restore to a previous state.

--BobBagwill

The more networked a computer gets, the more chances exist that a remote exploit will appear. Please make a list on the Wiki of any network-accessible services the OLPC will have, as they need to be scrutinised.

--SimosX

Some security features that commercial laptops lack that I would like are:

1. a reinforced grommet through the laptop in the center hinge area that would let you secure it to a flat surface
2. a reinforced eyelet for a cable lock
3. a standard threaded tripod socket

That would let you

• attach it to a wall
• easily secure them for a lab or cafe
• mount them on a swing arm or tilted table

-- BobBagwill 20:54, 5 June 2006 (EDT)

Encryption

Do you plan to encrypt the whole memory or does this cost too much power?

Encrypting all files on the filesystem is definitely going to cost many CPU cycles, as well as power. Most children will not have anything sensitive enough to need ubiquitous encryption, anyway. --SamatJain 15:48, 25 July 2006 (EDT)

signed code

Executing only signed code doesn't do any good unless the signed code is bullet proof. That seems highly unlikely.

I'm certain that at some point a critical mass will be reached where there are enough machines in use that there will be a general meltdown of the integrity of the system due to virus/trojan distribution. Given the slowness of these machines it will in effect render them useless.

Solving this problem requires solving problems that haven't been solved anywhere else yet. That would be a wonderful thing, but that task is perhaps 10 or 100 times bigger than the entire project of creating the OLPC hardware and software. It is in effect the tail wagging the dog.