CPS 463 Hardware Security – Targeted Exam Review

0.0(0)
Studied by 0 people
call kaiCall Kai
learnLearn
examPractice Test
spaced repetitionSpaced Repetition
heart puzzleMatch
flashcardsFlashcards
GameKnowt Play
Card Sorting

1/36

flashcard set

Earn XP

Description and Tags

Cache attacks — Compare Prime+Probe and Flush+Reload, what each must share, and what a slow probe or fast reload suggests. Side channels and covert channels — Distinguish unintended leakage from deliberate signaling. Compare SPA and DPA, calculate the difference of means, interpret a DPA peak, and explain how constant-time code reduces timing leakage. Physical attacks and fault injection — Classify non-invasive, semi-invasive, and invasive attacks by the access they require. Explain how clock or voltage glitches cause faults, how a fault can disrupt secure boot enforcement, and how VoltPillager targets the SVID link between the CPU and its voltage regulator. Hardware foundations and trust — Review the hardware root of trust, the IC lifecycle, what a PCB does, and how trust differs between third-party IP and chip fabrication. SoC test and debug security — Explain JTAG, boundary scan, and internal scan chains. Distinguish controllability from observability and identify how debug permissions protect assets such as keys and firmware. Hardware Trojans — Identify the trigger, payload, and effect, and explain why rare triggers can evade functional testing. DRAM refresh and RowHammer — Explain why DRAM needs refresh, how repeated aggressor-row activations can flip bits in a nearby victim row, and why a page-table bit flip can affect memory permissions.

37 Terms

1
New cards

What does a fast Flush+Reload result suggest?

The shared line was likely already cached when the observer reloaded it. The victim may have accessed it during the observation window.

2
New cards

What does a slow Prime+Probe result suggest?

Some of the observer's own lines were likely evicted. Victim activity in the same cache set is one possible cause.

3
New cards

What must be shared in each cache attack?

Flush+Reload needs the same physical cache line. Prime+Probe can use separate physical lines that compete for the same cache set.

4
New cards

How do you calculate the DPA difference of means?

Average each predicted-bit group separately at the same time sample. Then subtract in the order given by the question.

5
New cards

DPA practice: group 1 has 8 and 12. Group 0 has 10 and 16. What is mean₁ − mean₀?

Mean₁ is 10 and mean₀ is 13. The difference is −3.

6
New cards

What does a repeatable DPA peak support?

A peak near the relevant operation is evidence that the candidate key guess is consistent with the measured leakage.

7
New cards

What defense directly addresses timing leakage from a secret-dependent branch?

A constant-time implementation avoids control flow and memory accesses that depend on secret data.

8
New cards

Package intact during external power measurement

Non-invasive. The observer measures the device externally while its package remains intact.

9
New cards

Package opened for laser access to the die

Usually semi-invasive. The laser interacts with exposed circuitry without direct electrical contact with internal nodes.

10
New cards

A probe makes direct electrical contact with an internal chip net

Invasive microprobing. The probe directly contacts circuitry inside the chip.

11
New cards

Which interface does VoltPillager target?

Serial Voltage Identification or SVID. It carries voltage-control communication between the CPU and its voltage regulator.

12
New cards

VoltPillager: method vs. fault vs. impact

Board-level undervolting is the method. Incorrect computation is a possible fault. Compromising an enclave's security is the impact.

13
New cards

A supplier delivers an IP block with hidden malicious logic. What trust problem is this?

Third-party IP trust. The malicious logic is already present in the supplied design block.

14
New cards

A foundry inserts malicious logic during manufacture. What trust problem is this?

IC fabrication trust. The malicious modification enters during chip manufacturing.

15
New cards

Why does DRAM need refresh?

DRAM cells store bits as charge that gradually leaks. Refresh restores the charge before the data is lost.

16
New cards

Which row is repeatedly activated in RowHammer and which row can flip?

The aggressor row is repeatedly activated. A nearby victim row can experience bit flips even without a direct write to it.

17
New cards

Why can a RowHammer bit flip matter for security?

If it changes sensitive data such as a page-table permission then it can alter which code can access memory.

18
New cards
SPA vs. DPA
SPA directly inspects one or a few power traces for visible operation patterns. DPA compares many traces statistically to test key hypotheses.
19
New cards
Side channel vs. covert channel
A side channel reveals information through unintended leakage. A covert channel involves a cooperating sender deliberately encoding information.
20
New cards
Hardware Trojan trigger vs. payload vs. effect
The trigger activates the malicious logic. The payload performs the malicious action. The effect is the resulting leak or change in behavior.
21
New cards
Why can rare Trojan triggers evade functional testing?
Normal test inputs may never activate the payload. The chip can produce expected outputs throughout testing.
22
New cards
Scan-chain controllability vs. observability
Controllability lets a tester set internal state. Observability lets a tester read internal state. Unauthorized access can expose sensitive values such as keys.
23
New cards
Hardware root of trust
A trusted hardware-backed starting point that later security checks rely on.
24
New cards
IC lifecycle
Specification → design → fabrication → testing → assembly and packaging → customer.
25
New cards
How can a clock glitch cause a fault?
An early capture edge can sample a signal before it settles. The circuit may store an incorrect value.
26
New cards
What is the difference between a fault injection method and its security impact?
The method is the disturbance applied to the device. The impact is the security consequence. A voltage glitch can cause a control-flow error that leads to an access check being bypassed.
27
New cards
How do clock and voltage glitches cause timing faults?
A clock glitch can shorten the time between capture edges. Undervolting can slow circuit transitions. Either can cause a signal to arrive too late.
28
New cards
Secure boot verification vs. enforcement
Verification checks whether the next boot stage is trusted. Enforcement decides whether that stage may run. A fault can disrupt enforcement even when verification detects an invalid image.
29
New cards
What is JTAG and why does it matter for security?
JTAG is a test interface that provides access to functions such as boundary scan. Unauthorized access to test or debug features can expose or alter protected state.
30
New cards
Boundary scan vs. internal scan chain
Boundary scan observes or controls signals at chip I/O pins. An internal scan chain shifts state into or out of internal registers for testing.
31
New cards
Which assets can SoC test or debug access expose?
Cryptographic keys and device IDs and configuration data and proprietary firmware are examples of assets that may need protection.
32
New cards
How can an SoC preserve useful debug access while protecting secrets?
Authenticate the debugger and limit which operations and data its role can access. A field technician can get diagnostics while key registers stay restricted.
33
New cards
Why does board-level access matter for hardware security?
A person with physical access to a PCB may reach power connections or communication buses or debug headers. Those interfaces can expose signals or allow interference with device operation.
34
New cards
What are the two main jobs of a PCB?
It mechanically supports electronic components and electrically connects them through conductive traces. Multilayer boards route connections on several layers.
35
New cards
Why does SoC integration create a trust problem?
A single chip can combine blocks from multiple stakeholders. The integrator must assess whether supplied blocks contain hidden malicious logic.
36
New cards
How do physical attack classes generally differ in cost and evidence?
Non-invasive attacks typically cost less and leave little physical evidence. Semi-invasive attacks can leave signs of package opening. Invasive attacks tend to cost more and can damage the chip.
37
New cards
How does VoltPillager turn board-level access into a fault?
The case study shows that interfering with voltage-control commands on the SVID link can change CPU core voltage and cause incorrect computation.