346
4 Hardware Trojans in Microcircuits
Mirhoseyni [169] in continuation of [170]. The proposed infrastructure allows
carrying out simultaneous analysis of various third-party channels and using various
assessment methods, such as consumption current, current leakage, and delay.
The mathematical analysis of the measurement results here was based on the
characterization of logical elements and their subsequent statistical analysis. In this
case, a new objective function is set for the linear program, taking into account the
sub-modular nature of the problem. Obviously, the smaller the size of the analyzed
circuit, the stronger the impact of the Trojan on the third-party channel.
After characterization, for each logical circuit, the deviation of the measurement
results from the expected numerical values is calculated. Then a sensitivity analysis
is performed, which allows detecting possible malicious chips, although it is not
clear that the design of any Trojan determines its effect on third-party channels.
Some Trojans are more likely to affect the amount of power consumption, others—
the performance. The measurement results of various analyses (multimodal) are
combined to achieve a higher level of detection.
Experiments demonstrate that if Trojans are installed in regions with normal
sensitivity, the probability of such detection is 100%. The converse case is also true:
This method can be used to determine the areas where detecting Trojans is most
problematic.
Experts Lamech et al. [171] also assess the efficiency of combining the results of
analyses of various third-party channels. Unlike [169], however, they do not represent
a general model for combining the results obtained as a result of various third-party
channel analyses, but demonstrate that by combining the transient power and performance with the subsequent performance of the regression analysis, higher detection
rates can be achieved than when using each analysis separately. Experiments show
that the detection rate without calibration is up to 80%. After calibration, a detection level of up to 100% can be achieved. The multimodal analysis method will be
discussed in more detail in Chap. 6.
4.6.2.15 Increasing the Probability of Trojan Activation Due
to Additional Triggers
To increase the probability of state transitions in environments of microcircuits under
study in the process of functional tests, authoritative experts Salmani et al. [146, 183]
presented a new original approach, which involves inserting false scan triggers into
the original circuit. As a result, Trojans should fully or partially activate and have
a corresponding impact on third-party channels. For example, the logical elements
of a Trojan may turn on, and therefore during the third-party channel analysis, a
corresponding increase in the level of energy consumption can be observed. The
most important task of this method is to reduce the time for authorization of an IC,
without which its practical implementation would be problematic.
This method is performed as follows: first of all, the inclusion probability threshold
is determined, taking into account technical and economic conditions. After that, the
inclusion probability values of all network sections are determined, and the networks
4 Hardware Trojans in Microcircuits
Mirhoseyni [169] in continuation of [170]. The proposed infrastructure allows
carrying out simultaneous analysis of various third-party channels and using various
assessment methods, such as consumption current, current leakage, and delay.
The mathematical analysis of the measurement results here was based on the
characterization of logical elements and their subsequent statistical analysis. In this
case, a new objective function is set for the linear program, taking into account the
sub-modular nature of the problem. Obviously, the smaller the size of the analyzed
circuit, the stronger the impact of the Trojan on the third-party channel.
After characterization, for each logical circuit, the deviation of the measurement
results from the expected numerical values is calculated. Then a sensitivity analysis
is performed, which allows detecting possible malicious chips, although it is not
clear that the design of any Trojan determines its effect on third-party channels.
Some Trojans are more likely to affect the amount of power consumption, others—
the performance. The measurement results of various analyses (multimodal) are
combined to achieve a higher level of detection.
Experiments demonstrate that if Trojans are installed in regions with normal
sensitivity, the probability of such detection is 100%. The converse case is also true:
This method can be used to determine the areas where detecting Trojans is most
problematic.
Experts Lamech et al. [171] also assess the efficiency of combining the results of
analyses of various third-party channels. Unlike [169], however, they do not represent
a general model for combining the results obtained as a result of various third-party
channel analyses, but demonstrate that by combining the transient power and performance with the subsequent performance of the regression analysis, higher detection
rates can be achieved than when using each analysis separately. Experiments show
that the detection rate without calibration is up to 80%. After calibration, a detection level of up to 100% can be achieved. The multimodal analysis method will be
discussed in more detail in Chap. 6.
4.6.2.15 Increasing the Probability of Trojan Activation Due
to Additional Triggers
To increase the probability of state transitions in environments of microcircuits under
study in the process of functional tests, authoritative experts Salmani et al. [146, 183]
presented a new original approach, which involves inserting false scan triggers into
the original circuit. As a result, Trojans should fully or partially activate and have
a corresponding impact on third-party channels. For example, the logical elements
of a Trojan may turn on, and therefore during the third-party channel analysis, a
corresponding increase in the level of energy consumption can be observed. The
most important task of this method is to reduce the time for authorization of an IC,
without which its practical implementation would be problematic.
This method is performed as follows: first of all, the inclusion probability threshold
is determined, taking into account technical and economic conditions. After that, the
inclusion probability values of all network sections are determined, and the networks