In this course, we will study security and trust from the hardware perspective. Upon completing the course, students will understand the vulnerabilities in current digital system design flow and the physical attacks to these systems. They will learn that security starts from hardware design and be familiar with the tools and skills to build secure and trusted hardware.
Hardware Metering
Loading...
来自 University of Maryland, College Park 的课程
硬件安全
285 个评分
Explore Related Videos
At Coursera, you will find the best lectures in the world. Here are some of our personalized recommendations for you
从本节课中
Design Intellectual Property Protection
As a hardware designer or a company, you want to protect your design intellectual property (IP) from being misused (by users, competitors, silicon foundry, etc). We will cover how you can build such protection during the design process which can be used as an evidence to support law enforcement protection. You are expected to understand the basic digital logic design knowledge covered in week 1. We will use several NP-hard problems as examples to illustrate the concepts of IP protection. These problems (graph vertex coloring problem and graph partitioning problem) will be introduced in the lecture and you do not need to know the concept of NP-complete.
与讲师见面
Gang QuAssociate Professor
Electrical and Computer Engineering
Which takes financial increase in the complexity and
the cost of building a chip manufacturing line.
Most chip makers are fabless now.
In other words, today's design houses can not afford the in-house chip fabrication
cost and that they have to outsource this fabrication to foundries elsewhere.
This gives the foundries access to the details of the chips and the possibility
of overbuilding them without the authorization from the, the design house.
Integrated circuits metering, or
hardware metering, is an effective method to defend against this IP infringement.
Integrated circuit metering is a set of security protocols
that enable the design house to achieve post-fabrication control over their ICs.
IC metering techniques are developed to defend against IC overbuilding attacks.
As we have mentioned earlier, in the integrated circuit fabrication process,
these houses and foundries play asymmetric roles.
These houses have to disclose the details of their IPs and
pay the cost to build the mask in order to manufacture their, their chips.
Once the IP is released and
the mask is built, the design house will not have any control on them.
However, a dishonest foundry can utilize this information and
the facilities to reproduce the design with negligible cost.
This is known as the IP overbuilding.
Can we use digital watermarking techniques to protect a design,
designer's right in this case?
The answer is no.
Watermark protects the authorship of an intellectual property.
In IC overbuilding,
the foundry just reproduce more copies of the IP than required.
And then sells the extra copies to make profit.
The foundry will not attempt to claim the authorship of the IP.
How about digital fingerprint?
Fingerprint enables the trace of intellectual properties back to
the dishonest IP user who illegally distributed the IP.
When the overbuilt ICs carry a fingerprint,
the user who has this fingerprint will also become the victim of IC overbuilding.
This is because this innocent user will be,
now be considered as the source of these extra copies.
How integrated circuits metering or
hardware metering solves this problem then.
Let's first see how meters work in our daily life.
Why don't we talk about utility meters, like water meter or electricity meter?
The meters are installed and
maintained by the utility providers to monitor the users.
However, in the case of the IC overbuilding,
the foundry is the provider of the ICs.
And the design house who receive these ICs will be the user.
So what we need in, in our case is actually a reverse metering.
Where the user wants to monitor the provider.
The basic concept behind IC metering is to enable a unique tag to each copy
of the IC and make sure that the tag is under the control of the design house,
not the foundry.
Over the years, many different types of tags have been proposed and used for
hardware metering.
They can be categorized based on various criterias.
First, when the tag can only be used for
cheap identification, it is called a passive metering.
The tag in active metering can also do a lot of other functionalities,
such as enable the chip, disable the chip, or control the chip.
Active metering methods can be further classified as internal controlled or
external controlled based on whether the control is part of the design or not.
Intrinsic metering does not need help from additional hardware components,
or the modification of the design.
However, extrinsic metering methods do.
Depending whether the tag interacts with the chip's functionality or
not, we have non-functional metering and functional metering.
Finally, some tags can be reproduced and some cannot.
We call the latter unclonable tags, tags.
Serial number is perhaps the most popular and
one of the earliest of ways for device tagging.
A serial number can be physically indented on the device or
stored permanently in the memory.
This image shows a Trusted Platform Module in storage on a motherboard.
What we see here is, it has a couple of serial numbers.
For example here, and here.
These tags are passive, they are extrinsic because they do need additional,
like this piece of paper or trying to be physically indented on the, on the device.
And they are non-functional because they have nothing to do
with the functionality of the chip, and most important, they are reproducible.
You can use another tag or piece of paper to over, to cover this.
Or you can change the numbers here.
And it, it is the fact that this text can be reproduced
make them unsuitable to countermeasure foundry overbuilding.
The ICID tag was proposed around 2007.
It is based on the silicon manufacturer variation.
Because these variations occur during the chip fabrication process,
and it is generally believed that they are random and uncontrollable.
Therefore, ICID is considered as an unclonable tag,
and thus can be used for anti-overbuilding.
Depending whether the ID requires additional hardware components or
not, IC Identification, or ICID, can be, can be either intrinsic or extrinsic.
Let's first see an example of
an intrinsic unclonable ICID based on leakage current.
Consider a small subcircuit with four two-input NAND gate, G1,
G2, G3 and G4, with five input signals, x1, x2, x3, x4, and x5.
So here is the leakage of our ideal two-input NAND gate and
the different input combinations.
However, due to fabric, fabrication variation, the leakage on
the NAND2 gates will almost for sure be different from these ideal values.
For example, let's consider two ICs both have this two,
have the structure of four two-input NAND gates.
Their total leakage under four different input vectors are listed here in
this table, and we see a dramatic difference between their leakage values.
This leakage-based attack is intrinsic because it
does not require any extra hardware.
It is also unclonable because it is based on fabrication variations.
It is will be very hard for any foundries to reproduce this manufactured variations.
Now we show a functional passive tagging message.
Recall that in the graph vertex coloring problem, we want to assign each node
a color such that any pair of adjacent nodes will receive different colors.
In this case, the IP is a solution to the graph coloring problem.
Consider this graph and a coloring scheme, or an IP.
We observe that node H, for example, can also be colored with red or yellow,
because it has only two neighbors and they are both colored, colored with green.
And similarly, we observe that F can also receive a color, which is brown.
And finally, node E can also be colored by green instead of yellow.
So, therefore, we can create 12 different tags by
giving different colors to these three modes, H, E, and F.
This is a functional tag, because the tag,
which has the colors for node H, E, and F,
is part of the solution or part of the functionality of the IP.
Next, we see active integrated circuit metering methods.
Most of the existing active metering methods follow the following steps.
First, the design house will modify the design specification,
most likely in the early design stage, to insert a lock.
Then complete the design and ask the foundry to fabricate that design.
The foundry will fabricate, and then, maybe, overbuild the chips.
Due to manufacturer variations, each chip with each IC, will have a unique and
unclonable identifier that can be used to authenticate the chip.
In the active IC metering scheme,
the design house can use this unique ID to actively meter or monitor the IC.
For example, by enabling, disabling, locking, or unlocking the ICs.
Here is an example of an internal active IC metering approach.
The design house introduces additional flip-flops to
build what is called a boosted finite state machine.
Notice that each added flip-flop will double the number of states in
the original finite state machine.
Therefore, when we power up the chip and
use the path data to select a random power up state.
Most likely this state does not belong to any of the original final state machine.
But for the chip to function properly, it has to start from certain initial state.
The design house will provide a chip with power up states specific input
sequence to direct the random power up state to the initial state,
such that the chip can work properly.
Without this correct input sequence, the, the,
the IC will not be, it will be useless.
This active IC metering approach uses internal finite state machine
to control chips, to, to control the monitoring of chip.
The additional flip-flops make it extr extrinsic.
The use of PUF ensures that it is unclonable and
thus can be used to countermeasure chip overbuilding attack.
Finally, we show an external active IC metering approach.
In this approach, the design house will add control signals and
logical units such as exclusive or gate to non-critical parts of the design.
Unless all the control signals receive the proper values,
the chip or the fabricated ICs will be locked.
The design house will then provide an external key to unlock each of
the ICs based on some asymmetric cryptographic primitives, for
example, the public key infrastructure.
In this case, we see this is active metering scheme.
Because it does involve the interaction between the,
the design house and each fabricated ICs.
It is an external method because the signal comes from
the external value provided by design house.
It is extrinsic because we do introduce additional control signals and
logical gates.
And finally, it is also unclonable,
which means that it can also be used to countermeasure circuit overbuilding.
