SecureEasySetup offers one-push security, but does it deter malicious attacks? A few days ago, I wrote about Broadcom's new SecureEasySetup security offering, which allows a home user to setup and configure WPA Personal by pressing a button (software or hardware) on an access point and any client device, whether a computer adapter or a consumer-electronics device. The system is pretty slick and has a great degree of security and ease of use.
But in contrasting SES with a similar offering from Atheros, JumpStart for Wireless, both of which were announced during CES last week, I noticed what I thought was a gap in the security during a period of contention for SES. Atheros's system requires that a user enter a password when setting up security, but the password is out of band--not revealed over an insecure network--thus allowing a simple password to create and distribute a robust WPA encryption key. (For more details on both systems, consult this in-depth article contrasting the two.)
SES seems to lack true out-of-band security. That is, once the button is pressed to initiate a key exchange on an access point, any client within range can jump in and receive the key if they slip in the queue during the two-minute period of contention before the user pushes the button or initiates SES on the client device they want to add. (After two minutes, the button needs to be pushed on the AP again to restart SES; Broadcom estimates most users will start the key exchange in a matter of a few seconds on the client device.)
I spoke today with David Cohen, who has been deeply involved with WPA at the Wi-Fi Alliance in his role as senior product marketing manager at Broadcom. Cohen noted that there are methods by which an out-of-band element is layered on top of SES to prevent malicious clients or putative cracker software from becoming valid clients on a network in which security keys are managed by SES.
The first is that client devices will be expected to provide a warning when another client slips in the queue during the two minutes from when the access point is activated to watch for a client's request before the desired client. For instance, if I push the AP button, walk over to my computer, and click a software button, the Broadcom system includes a monitoring element that will allow the client software to signal that my client didn't successfully gain a key because another client gained access. "You'd have to be realistically under deliberate attack and then you'd get notification," Cohen said.
On a consumer-electronics device, Cohen said the device could either signal this problem through a display--such as an LCD or TV--or for non-display devices it could flash a red LED rapidly. Cohen emphasized that Broadcom is providing a framework and that individual manufacturers will choose how to integrate SES into their products, including this sort of feedback to the user. But Broadcom considers this notification significant because it removes edge conditions for security.
In this case, if a user then wanted to change the key for their network, they would hold down the access point SES button for at least five seconds or choose an option via software for APs without a hardware button. This triggers the AP to generate a new key, and now the user can re-establish keys for their devices. If the problem recurs, there's a second option for client devices that have both Ethernet and Wi-Fi interfaces.
The second option provides total out-of-band security: you can operate SES over a wired network. Plug your laptop into an Ethernet network or a port on the access point, push a button on the AP, and now push a button on a client. SES handles the transaction out of the Wi-Fi band--in the wired band, which is ostensibly far more secure on a home network--and then the device can be used on the Wi-Fi network. (Atheros's JumpStart requires that the AP is configured with a password via an Ethernet network initially, which allows clients to join over Wi-Fi using a password that was protected out of band.)
Cohen emphasized that my scenario of a malicious attack--and accidental one is extremely unlikely--is an edge case. "The global problem we're trying to solve is over 80 percent of the networks out there are wide open," Cohen said. "Hackers are going to jump on those open networks. We want to bring that number down."
Cohen's response to my concern answers a fairly wide swath of my issues about a lack of out-of-band confirmation, but it still leaves open what is more realistically a denial-of-service (DoS) attack against SES rather than a security hole. Because of the notification element that's part of SES, a malicious client won't go undetected. But it could make it impossible for a user to set up SES. (They can always fall back on manual WPA setup at that point, of course, or employ the "nuclear" trigger: one finger on the AP button and one finger on the client button.)
If malicious client software were written so that it constantly scanned for SES transactions and instantly leapt on any AP that's looking for a partner to dance with, this could block a user's ability to ever set their network up with SES. But I can see a small improvement that would prevent this. SES could cause an AP, after having SES reset to produce a new key, to reject the last successful SES transaction's MAC address or other radio and network characteristics.
In that scenario, the rogue client may be constantly scanning, but it's locked out during the next cycle--or even permanently. (Of course, MAC addresses can be spoofed or generated randomly, so I may not be thinking big enough.) Because this would only be triggered when a user reset the SES key on the AP and only after two clients contended and one lost, this wouldn't be trigger in normal configuration circumstances. This would be a firmware change, so it's doesn't require a radical rethink on the part of Broadcom or its partners.
SES is almost all the way there, and it does still have an advantage of ease of use over Atheros's JumpStart. JumpStart clearly is trying to combine the best of out-of-band security while still using the same band for securing a network, but it's interface issues--having to enter a password--definitely bumps its complexity higher. Broadcom would like to keep the margin of attack small, but if the wrong software were to appear, users might find themselves slightly stymied and turned to manufacturers for help in sorting out just what's happening on their network.