Tamper evidence for verifier devices

How it works

Verification hardware placed in a facility run by the party being checked needs protection against physical interference ¹². The MIRI system overview relies on monitoring of the facility and on occasional random inspections of analog components and anti-tamper seals ¹². Obermaier and Immler review enclosures from battery-backed monitoring to designs based on physical unclonable functions (PUFs) ³. Published approaches fall into four groups:

• Tamper-respondent modules. The IBM 4765 coprocessor uses a "protective, tamper-respondent matrix to monitor for intrusion and adverse physical conditions" ¹. On a hard tamper, it actively erases its core secrets within microseconds ¹.
• Batteryless tamper-resistant covers. Immler et al. wrap the protected system in a flexible circuit-board cover with a fine electrode mesh ⁴. A key is derived from the mesh's capacitances, which act as a PUF, and decrypts the system's sensitive data only if the cover is intact ⁴.
• System-level sensing. Anti-Tamper Radio monitors how radio waves propagate inside a metal case, so inserted objects change the measured response ⁵. ImpedanceVerif uses on-chip network analyzers on FPGAs to detect changes in a board's power distribution network, "without any modifications to the system" ⁸.
• Tamper-indicating enclosures. In safeguards and arms control, these leave physical evidence of attack ⁹. They cover the case where an adversary bypasses a sealed opening, for example by drilling through a side ⁹.

In AI proposals, flexHEG pairs a guarantee processor with "a secure enclosure providing physical tamper protection" ¹⁴; see Hardware-enabled guarantees (flexHEG) and guarantee processors. The AI 2040 plan lists tamper-evident enclosures among physical security measures ¹³. The MIRI overview notes that standards exist for secure facilities (SCIFs) ¹². What is less established, it says, is "what defenses can be retrofitted at a massive scale to prevent bypassing of network taps" ¹².

AI Security Forum's sealed-cluster proposal seeks to avoid maintenance access through redundant components, overprovisioning and conservative operating settings ¹⁷. Its retrofit sketch proposes sensors and a controller that would shut down a server tray after detecting tampering ¹⁶.

The Oxford Martin AI Governance Initiative's report sketches verifying hardware inside a sealed shipping container for use as a mobile data centre, then maintaining digital contact intended to conceal its location ¹⁵. Location leakage and the security of hardware roots of trust remain unresolved in this speculative proposal ¹⁵.

What it establishes

Enclosures can show that a device was physically accessed ⁹, or can make a breach destroy the keys or data an attacker wants ^{1 4}. They protect the integrity of other mechanisms, such as taps (Network taps and certifiers) and bandwidth shapers (Bandwidth limits and compartmentalization), rather than proving a claim on their own ¹². Johnston argues that "a seal is no better than the protocols for using it" ¹¹.

Threat model

• Cover. Immler et al. "assume penetrations to be at least 300µm in diameter" ⁴. They deem cover removal impractical and do not assess chemical solvents ⁴.
• Radio sensing. Anti-Tamper Radio defines success for the attacker as inserting a needle into a sensitive region without detection ⁵.
• AI facilities. The host controls the facility, so the MIRI system overview combines enclosures with continuous monitoring and random inspection ¹².