CWE-1300 is about information that leaves a device through the physical world instead of the data path. The device does not contain sufficient protection mechanisms to prevent physical side channels from exposing sensitive information, through patterns in observable phenomena such as variations in power consumption, electromagnetic emissions (EME) or acoustic emissions.
If your threat model stops at the network or the USB port, this is the part it leaves out.
How it happens
MITRE describes an adversary who monitors and measures physical phenomena to detect patterns and make inferences, even when extracting the information in the digital domain is impossible. Physical side channels have been studied for decades, mostly in the context of breaking implementations of cryptographic algorithms and other security features.
They can be observed by someone with physical access, or by using a tool in close proximity. If the attacker can watch the hardware operate and correlate its data processing with power, EME and acoustic measurements, they might recover secret keys and data.
The record gives two demonstrative examples.
First, a device that checks a passcode to unlock the screen. As each character of the PIN is entered, a correct character produces one current pulse shape and an incorrect character a different one. An attacker watching with an oscilloscope can confirm the first character, move to the next, and so on. That is far more efficient than brute forcing every combination. MITRE's fix is to accumulate the PIN in a register and compare it all at once at the end, or to change the comparison components so the current pulse shape is the same whether or not the character is correct.
Second, CVE-2021-3011, which affects certain microcontrollers. The Google Titan Security Key uses an internal secret key for two-factor authentication. The local extraction method is plugging the key into a USB port and using electromagnetic sniffing tools and computers. If an adversary learned the secret key and the algorithm, they could duplicate the function and masquerade as the legitimate user. MITRE says the manufacturer could shield the circuitry, add randomized delays, use indirect calculations with random values, or randomly order calculations, and could combine these approaches.
Other observed examples from the record:
- CVE-2022-35888: power side channels leak secret information from a processor.
- CVE-2019-14353: a crypto hardware wallet's power consumption relates to the total number of pixels illuminated, creating a side channel on the USB connection that lets attackers determine secrets shown on the display, such as the PIN.
- CVE-2020-27211: a chain involving a microcontroller system-on-chip that uses a flash register value to set product protection state and lacks protection against fault injection.
What an attacker gains
The affected scope is confidentiality, and the consequences listed are reading memory and reading application data. This is disclosure of keys, PINs and secrets, which usually leads straight to impersonation.
The attacker needs physical access or close proximity, so the risk is highest for devices that are carried, lost, stolen, shipped or placed in untrusted locations: tokens, wallets, phones and embedded gear.
How to find it
MITRE lists three manual analysis methods.
- Leakage detection tests, such as the procedure in the Test Vector Leakage Assessment (TVLA) requirements for AES. TVLA is the basis for ISO standard 17825, and a separate methodology is also referenced. The record cautions against relying only on this kind of testing.
- Post-silicon, full side-channel attacks as penetration testing, covering as many known leakage models as possible against test code, rated moderate.
- Pre-silicon work. TVLA-style methods can be used post-silicon, but models of power consumption or other physical emanations can be built from information present at various design stages before fabrication, and the same tests or known attacks can be run against those models.
In practice that means testing early on models and again on real parts, and treating a clean TVLA result as one input rather than a guarantee.
How to fix it
MITRE lists two mitigations.
- At architecture and design, apply blinding or masking techniques to implementations of cryptographic algorithms, so the physical signal is decoupled from the secret.
- At implementation, add shielding or tamper-resistant protections to the device to make it harder to obtain measurements of the side channel.
From the examples you can add design habits that follow naturally: avoid comparisons whose physical footprint differs depending on whether the input was right, make operation timing and ordering less predictable where the design allows, and do not let displayed or processed secrets drive a visible pattern on shared lines.
Quick check
Why is character by character PIN checking a problem?
Each correct character produces a different current pulse than a wrong one, so an attacker can confirm one character at a time instead of guessing the whole PIN.
Does an attacker need network access?
No. MITRE describes observation by someone with physical access or close proximity, using power, EME or acoustic measurement.
What do blinding and masking do?
They change how a cryptographic algorithm is implemented so measurable physical patterns no longer line up with the secret data.
