CVE-2022-39218

The JS Compute Runtime for Fastly's Compute@Edge platform provides the environment JavaScript is executed in when using the Compute@Edge JavaScript SDK. In versions prior to 0.5.3, the Math.random and crypto.getRandomValues methods fail to use sufficiently random values. The initial value to seed the PRNG (pseudorandom number generator) is baked-in to the final WebAssembly module, making the sequence of random values for that specific WebAssembly module predictable. An attacker can use the fixed seed to predict random numbers generated by these functions and bypass cryptographic security controls, for example to disclose sensitive data encrypted by functions that use these generators. The problem has been patched in version 0.5.3. No known workarounds exist.

The JS Compute Runtime for Fastly's Compute@Edge platform provides the environment JavaScript is executed in when using the Compute@Edge JavaScript SDK. In versions prior to 0.5.3, the Math.random and crypto.getRandomValues methods fail to use sufficiently random values. The initial value to seed the PRNG (pseudorandom number generator) is baked-in to the final WebAssembly module, making the sequence of random values for that specific WebAssembly module predictable. An attacker can use the fixed seed to predict random numbers generated by these functions and bypass cryptographic security controls, for example to disclose sensitive data encrypted by functions that use these generators. The problem has been patched in version 0.5.3. No known workarounds exist.

Impact

Math.random and crypto.getRandomValues methods failed to use sufficiently random values. The initial value to seed the CSPRNG (cryptographically secure pseudorandom number generator) was baked-in to the final WebAssembly module meaning the sequence of numbers generated was predictable for that specific WebAssembly module. An attacker with access to that same WebAssembly module that calls the affected methods could use the fixed seed to predict random numbers generated by these functions. This information could be used to bypass cryptographic security controls, for example to disclose sensitive data encrypted by functions that use these generators.

Patches

The problem has been fixed in version 0.5.3.

Corrected Math.random and crypto.getRandomValues methods to always use sufficiently random values. The previous versions would use a CSPRNG (cryptographically secure pseudorandom number generator) which we would seed with a random value. However, due to our use of Wizer, the initial value to seed the CSPRNG was baked-in to the final WebAssembly module meaning the sequence of numbers generated was predictable for that specific WebAssembly module. The new implementations of both Math.random and crypto.getRandomValues do not use a CSPRNG and instead pull random values from WASI (WebAssembly System Interface) libc’s random_get function, which is always a sufficiently random value.

Workarounds

There are no workarounds, you must upgrade to version 0.5.3 or later.

The JS Compute Runtime for Fastly's Compute@Edge platform provides the environment JavaScript is executed in when using the Compute@Edge JavaScript SDK. In versions prior to 0.5.3, the Math.random and crypto.getRandomValues methods fail to use sufficiently random values. The initial value to seed the PRNG (pseudorandom number generator) is baked-in to the final WebAssembly module, making the sequence of random values for that specific WebAssembly module predictable. An attacker can use the fixed seed to predict random numbers generated by these functions and bypass cryptographic security controls, for example to disclose sensitive data encrypted by functions that use these generators. The problem has been patched in version 0.5.3. No known workarounds exist.

The JS Compute Runtime for Fastly's Compute@Edge platform provides the environment JavaScript is executed in when using the Compute@Edge JavaScript SDK. In versions prior to 0.5.3, the Math.random and crypto.getRandomValues methods fail to use sufficiently random values. The initial value to seed the PRNG (pseudorandom number generator) is baked-in to the final WebAssembly module, making the sequence of random values for that specific WebAssembly module predictable. An attacker can use the fixed seed to predict random numbers generated by these functions and bypass cryptographic security controls, for example to disclose sensitive data encrypted by functions that use these generators. The problem has been patched in version 0.5.3. No known workarounds exist.

Impact

Math.random and crypto.getRandomValues methods failed to use sufficiently random values. The initial value to seed the CSPRNG (cryptographically secure pseudorandom number generator) was baked-in to the final WebAssembly module meaning the sequence of numbers generated was predictable for that specific WebAssembly module. An attacker with access to that same WebAssembly module that calls the affected methods could use the fixed seed to predict random numbers generated by these functions. This information could be used to bypass cryptographic security controls, for example to disclose sensitive data encrypted by functions that use these generators.

Patches

The problem has been fixed in version 0.5.3.

Corrected Math.random and crypto.getRandomValues methods to always use sufficiently random values. The previous versions would use a CSPRNG (cryptographically secure pseudorandom number generator) which we would seed with a random value. However, due to our use of Wizer, the initial value to seed the CSPRNG was baked-in to the final WebAssembly module meaning the sequence of numbers generated was predictable for that specific WebAssembly module. The new implementations of both Math.random and crypto.getRandomValues do not use a CSPRNG and instead pull random values from WASI (WebAssembly System Interface) libc’s random_get function, which is always a sufficiently random value.

Workarounds

There are no workarounds, you must upgrade to version 0.5.3 or later.