Alexa versus Alexa: Controlling Smart Speakers by Self-Issuing Voice Commands

Introduction to the show: ident: Security Radio. Generated commentary on the latest security and cryptography papers.

Nadia: Hi, I'm Nadia.

Elias: And I'm Elias, here with Priya.

Nadia: We're looking at the paper: "Alexa versus Alexa".

Elias: Here's the gist of the paper. The gist: We present Alexa versus Alexa (AvA), a novel attack that leverages audio files containing voice commands and audio reproduction methods in an offensive fashion, to gain control of Amazon Echo devices for a prolonged amount of time How it works The AvA attack leverages the fact that Alexa running on an Echo device correctly interprets voice commands originated from audio files even when they are played by the device itself, which is referred to as a command self-issue vulnerability This vulnerability removes the necessity of having a rogue speaker in proximity of the victim’s Echo, a constraint that many attacks share.

Nadia: With AvA, an attacker can self-issue any permissible command to Echo, controlling it on behalf of the legitimate user.

Elias: The attack starts when the Echo device begins streaming an audio file that contains voice commands.

Nadia: This can be done, for example, by opening a malicious skill that makes the Echo device tune in a radio station that streams such voice commands.

Paper summary: Nadia: So to recap, this paper presents Alexa versus Alexa as a novel attack using audio files and reproduction methods to gain prolonged control over Amazon Echo devices.

Elias: The central thesis is that AvA leverages a command self-issue vulnerability where Alexa interprets commands correctly even when the audio is played back by the device itself.

Priya: This bypasses the need for a rogue speaker nearby, which cuts down on one of the main requirements for many other types of attacks.

Nadia: With this method, an attacker can self-issue any command to Echo and control it as if they were the legitimate user.

Elias: They verified that via AvA, attackers can control smart appliances in a household, buy unwanted items on Amazon, and tamper with linked calendars.

Priya: The research also discovered two specific Echo vulnerabilities: Full Volume and Break Tag Chain.

Nadia: The Full Volume vulnerability increases the self-issue command recognition rate by doubling it on average, allowing for more self-issue commands.

Elias: The Break Tag Chain vulnerability increases the time a skill can run without user interaction from eight seconds to over an hour.

Priya: Essentially, these combined vulnerabilities allow an adversary to keep control of the device for a prolonged amount of time with high success rates.

Conclusion: Nadia: So, looking at Alexa versus Alexa, the authors are showing that exploiting internal audio handling mechanisms can lead to serious device compromise through self-issued commands.

Elias: The implication is that security teams need to be paying attention to how devices process audio streams internally for these kinds of vulnerabilities.

Priya: For the average person, it means being aware of the risks associated with using skills that involve playing audio on their Echo and understanding what those skills are allowed to do.

Nadia: It's about understanding that even seemingly harmless features can have exploitable pathways if they allow for self-issuance.

Elias: The authors identified these vulnerabilities, and while Amazon rated them Medium severity, the fact that they found ways to maintain control for hours is what makes this specific research interesting.

Priya: So in short, it highlights that the risk isn't just about external access; it’s about vulnerabilities built into how the device handles its own audio output.