348
4 Hardware Trojans in Microcircuits
sequence. After n of processed bits, a false load is used to avoid the activation of a
Trojan.
4.6.2.17 Using Ring Oscillators for Detecting Trojans
Researchers Zhang and Tehranipur in work [190] use a network of ring oscillators
to detect the fact of installed Trojans. Such a ring oscillator is a simple circuit for
generating oscillations, consisting of an odd number of similar elements.
Here, the principle of detecting Trojans was based on the fact that physical parameters affect the frequency of the ring oscillator. Accordingly, the oscillator frequency
also depends on the supply voltage V. If the value of V drops, the propagation delay
of elements increases. This, in turn, means that the delay of the entire ring oscillator
and the duration of its cycle increase, which is equivalent to a drop in frequency.
It is clear that a drop in the V value occurs if the element begins to consume
current. If CMOS is to be used, a drop occurs with each switch of the transistor, that
is, in each case of a change in its state. If a Trojan is installed in the circuit, adjacent
ring oscillators will register a more significant drop in V DD and frequency compared
to a system that does not contain Trojans.
It has to be said that in order to achieve the highest possible coverage, ring oscillators were installed over the entire surface of the microcircuit. With the help of statistical methods, due to which the frequencies of the built-in ring oscillators are assessed,
the detection probability of 100% is achieved here. Testing a method using FPGAtype microcircuits makes it possible to achieve an accuracy in the range of 80–100%.
The efficiency of this approach in dealing with direct attacks is considered extremely
high, since the proposed manipulations have a direct impact on the frequency of operation of ring oscillators, thereby revealing themselves when performing functional
tests.
So, this section provides a brief overview of the main known methods for detecting
hardware Trojans in microcircuits. In Chap. 5, specially devoted to this problem, the
mechanisms for the implementation of both the main methods listed above and a
number of others, including such exotic (but not less efficient) methods as methods
based on the classical mathematical theory of card games and many others, will be
considered in more detail.
4.7 Case Study of the Development and Implementation
of a Hardware Trojan
Here we take a closer look at the results of the implementation of another project on
hardware Trojans, which placed first at the already mentioned Embedded Systems
Challenge at the 2008 Computer Security Awareness Week (CSAW) conference at
4 Hardware Trojans in Microcircuits
sequence. After n of processed bits, a false load is used to avoid the activation of a
Trojan.
4.6.2.17 Using Ring Oscillators for Detecting Trojans
Researchers Zhang and Tehranipur in work [190] use a network of ring oscillators
to detect the fact of installed Trojans. Such a ring oscillator is a simple circuit for
generating oscillations, consisting of an odd number of similar elements.
Here, the principle of detecting Trojans was based on the fact that physical parameters affect the frequency of the ring oscillator. Accordingly, the oscillator frequency
also depends on the supply voltage V. If the value of V drops, the propagation delay
of elements increases. This, in turn, means that the delay of the entire ring oscillator
and the duration of its cycle increase, which is equivalent to a drop in frequency.
It is clear that a drop in the V value occurs if the element begins to consume
current. If CMOS is to be used, a drop occurs with each switch of the transistor, that
is, in each case of a change in its state. If a Trojan is installed in the circuit, adjacent
ring oscillators will register a more significant drop in V DD and frequency compared
to a system that does not contain Trojans.
It has to be said that in order to achieve the highest possible coverage, ring oscillators were installed over the entire surface of the microcircuit. With the help of statistical methods, due to which the frequencies of the built-in ring oscillators are assessed,
the detection probability of 100% is achieved here. Testing a method using FPGAtype microcircuits makes it possible to achieve an accuracy in the range of 80–100%.
The efficiency of this approach in dealing with direct attacks is considered extremely
high, since the proposed manipulations have a direct impact on the frequency of operation of ring oscillators, thereby revealing themselves when performing functional
tests.
So, this section provides a brief overview of the main known methods for detecting
hardware Trojans in microcircuits. In Chap. 5, specially devoted to this problem, the
mechanisms for the implementation of both the main methods listed above and a
number of others, including such exotic (but not less efficient) methods as methods
based on the classical mathematical theory of card games and many others, will be
considered in more detail.
4.7 Case Study of the Development and Implementation
of a Hardware Trojan
Here we take a closer look at the results of the implementation of another project on
hardware Trojans, which placed first at the already mentioned Embedded Systems
Challenge at the 2008 Computer Security Awareness Week (CSAW) conference at