712
7 Countermeasures Against Hardware Trojans
extrinsic. On the contrary, intrinsic (internal) passive metering methods require no
additional components or modifications of the design. A big advantage of all intrinsic
identification methods consists in the fact that they don’t involve introduction of
any additional components and can be easily implemented on existing microcircuit
designs. Intrinsic identification can be based on various digital or analog values
caused by so-called random physical disturbances of order in events.
An example of such intrinsic digital identification is a weak PUF placed in the
structure of memory circuits of the SRAM type, where the existing SRAM memory
cells inside the FPGA are used as uncloneable IDs [123]. An example of intrinsic
analog identification applicable both to ASICs and FPGAs is the method of extraction
of a special signature determined by chaotic changes in the microcircuit production
process parameters [97, 120]. The extracted signatures can be used as microcircuit
fingerprints. While earlier works were focused on analyzing and using signatures of
dynamic characteristics [121], the latest papers demonstrated that it is also possible
to use such signatures from other side channels as measurements of IDDQ and IDDT
dynamic currents [117].
Modern methods of active hardware metering, in addition to the possibility of
ensuring unique identification of a semiconductor device and/or its functionality,
can act remotely. It is hard to believe, but the method actually allows the expert to
solve reverse tasks as well—remotely permit functioning, control, or even block a
manufactured circuit.
The first method of such active hardware metering of ASIC chips known to the
authors was suggested in [97]. Speaking in literary terms instead of technical ones,
we can say that active metering significantly enriched the sphere of classic functional metering, revealing states and structures that can only be available to a highly
qualified microcircuit developer. Results of the latest studies in the field of active
metering demonstrate that the original method suggested in [97] can be practically
implemented so as to be provably safe, transparently demonstrating mathematical
transformation of provably tricky family of generalized point functions, as mathematicians would put it [125]. Since logic states and various transactions used for
systematic control (including interlock and release operations) of chips are integrated
within the limits of functional specification of the design, specialists refer to this type
of metering as internal active hardware metering.
Since the first ideas of active hardware metering were suggested in [97], many
researchers have proposed several other equally interesting methods. For example, in
addition to the methods of internal active hardware metering, which are often used to
embed interlock modes only by means of design modification (sequential or combinational circuitry solutions) [94], other effective methods of active hardware metering
have been suggested, which are based on including external cryptographic circuits
[125, 133–135]. The authors of [116] refer to this type of active metering as external
active hardware metering. Both intrinsic and extrinsic methods of active hardware
metering usually implement technical solution of random identifier in digital form,
which may or may not be reproducible. For example, regular fusible links can theoretically be used as such random digital identifier. Let us note that the process of
blowing fuse links in chips, as a rule, is only implemented in serial production and
7 Countermeasures Against Hardware Trojans
extrinsic. On the contrary, intrinsic (internal) passive metering methods require no
additional components or modifications of the design. A big advantage of all intrinsic
identification methods consists in the fact that they don’t involve introduction of
any additional components and can be easily implemented on existing microcircuit
designs. Intrinsic identification can be based on various digital or analog values
caused by so-called random physical disturbances of order in events.
An example of such intrinsic digital identification is a weak PUF placed in the
structure of memory circuits of the SRAM type, where the existing SRAM memory
cells inside the FPGA are used as uncloneable IDs [123]. An example of intrinsic
analog identification applicable both to ASICs and FPGAs is the method of extraction
of a special signature determined by chaotic changes in the microcircuit production
process parameters [97, 120]. The extracted signatures can be used as microcircuit
fingerprints. While earlier works were focused on analyzing and using signatures of
dynamic characteristics [121], the latest papers demonstrated that it is also possible
to use such signatures from other side channels as measurements of IDDQ and IDDT
dynamic currents [117].
Modern methods of active hardware metering, in addition to the possibility of
ensuring unique identification of a semiconductor device and/or its functionality,
can act remotely. It is hard to believe, but the method actually allows the expert to
solve reverse tasks as well—remotely permit functioning, control, or even block a
manufactured circuit.
The first method of such active hardware metering of ASIC chips known to the
authors was suggested in [97]. Speaking in literary terms instead of technical ones,
we can say that active metering significantly enriched the sphere of classic functional metering, revealing states and structures that can only be available to a highly
qualified microcircuit developer. Results of the latest studies in the field of active
metering demonstrate that the original method suggested in [97] can be practically
implemented so as to be provably safe, transparently demonstrating mathematical
transformation of provably tricky family of generalized point functions, as mathematicians would put it [125]. Since logic states and various transactions used for
systematic control (including interlock and release operations) of chips are integrated
within the limits of functional specification of the design, specialists refer to this type
of metering as internal active hardware metering.
Since the first ideas of active hardware metering were suggested in [97], many
researchers have proposed several other equally interesting methods. For example, in
addition to the methods of internal active hardware metering, which are often used to
embed interlock modes only by means of design modification (sequential or combinational circuitry solutions) [94], other effective methods of active hardware metering
have been suggested, which are based on including external cryptographic circuits
[125, 133–135]. The authors of [116] refer to this type of active metering as external
active hardware metering. Both intrinsic and extrinsic methods of active hardware
metering usually implement technical solution of random identifier in digital form,
which may or may not be reproducible. For example, regular fusible links can theoretically be used as such random digital identifier. Let us note that the process of
blowing fuse links in chips, as a rule, is only implemented in serial production and