Form 1

PROGETTO DI RICERCA / PROJECT

Cognome e nome del candidato

SUTAPALLI KRISHNA CHAITANYA
Applicant’s Full Name

Corso di dottorato
PhD IN COMPUTER SCIENCE
PhD

1) Titolo del progetto / Project title

Spintronics based MRAM for Security using PUF (Physical Unclonable

Function) and PGP (Pretty Good Privacy)

2) Sommario / Abstract

With the advent of high-performance computing and mobile devices, enormous information is
generated and stored. This is made possible due to the growth in the storage and memory
technologies. Hard disk drives (HDD) with a capacity of 10 MB were sold for $5300 in the 1980s.
However, HDDs with 12 TB capacities (a million times larger capacity) are available by 2017. Even
mobile gadgets have a dynamic random-access memory (DRAM) capacity of 4 GB.
Despite such a growth, dynamic RAMs (DRAMs) which are based on semiconductor technologies
face limitations in maintaining a significant growth rate. The charge leakage that occurs when the
device sizes are scaled down causes an increase in power consumption. A non-volatile memory would
not consume much more static power, in contrast with dynamic RAMs. Due to potentially lower
power consumption and instant-on capability, a search for non-volatile memory technologies has been
actively pursued for and the answer is MRAM.
Magneto resistive random access memory (MRAM) is a non-volatile and non-destructive read out
memory, which is based on magnetic anisotropy energy to retain information and the principle of
magneto resistance to retrieve information. In Recent times, MRAM was released as a memory for
some niche applications. It is believed that MRAM could possibly replace Static Random-Access
Memory (SRAM) and DRAM as technology.

3) Obiettivi e rilevanza dei risultati ottenibili nel contesto dello stato dell’arte / Project
aims and their relevance in the context of the state of the art

The security mechanism utilizes an encryption key which is stored in, or is generated from
information stored in, a non-volatile memory module, such as a magneto resistive random-access

1
memory (MRAM) module, integrated in the data storage system. Data stored in the MRAM module
can be erased, written over, or otherwise rendered unrecoverable by exposing the MRAM module to a
suitable electromagnetic field. For example, data stored in the MRAM module can be erased by
degaussing. Such as by using a degausser configured to erase data stored in magnetic media.
As another example, the data storage system can comprise a magnetic field generator that can be
activated (e.g., by supplying external power) to erase data stored in the MRAM module.
Advantageously, by erasing data stored in the MRAM module, access to encrypted data stored in the
data storage system can be prevented even in cases when the data storage system has failed or has
otherwise been rendered in operative or nonfunctional.
In other cases, the data storage system is configured to erase from the MRAM module or otherwise
render unrecoverable (e.g., written over) the encryption key or information from which the encryption
key is generated. For example, this can be performed in response to a command (e.g., a vendor
specific command that is not part of Standard command interface between a host system and the data
storage system) received from a host system.
Once the key or information from which the key is generated is erased, encrypted data stored in the
data storage system cannot be decrypted, and access is prevented. As a result, the data storage system
is sanitized.

4) Descrizione del progetto / project description

Building Blocks:
 MRAM
 STT based MRAM
 Physical Unclonable Function
 Pretty Good Privacy

 MRAM:

A Magneto resistance random-access memory (MRAM) stores information as a magnetic orientation

rather than as an electrical charge. This immediately provides a much higher reading and writing
performance that is much closer to DRAM speeds than flash because bits are read by testing with
voltage, not current, and written with a small current boost, not a huge charge.
A typical MRAM cell has a transistor and a Magneto resistive element, quite like a DRAM, which
contains a transistor and a capacitor. While the charge stored in the capacitor of a DRAM defines its
memory state, the resistance of the Magneto resistive element determines 1 and 0 states. A transistor
for every MRAM cell is required, as the absolute difference between the resistances and hence, the
voltage of two states is not high enough to function without a transistor. Moreover, the transistor also
provides the current required for the write operation.

2
  STT based MRAM:

STT stands for Spin-torque transfer, STT-MRAM is based on letting a current ow through the MTJ
device itself to realize the switching of the magnetization of the free layer. This means that a current
owing through the device can both sense the resistance (for reading operation) and program (write
operation) the memory. The current is seen as a set of spin-up and spin-down electrons. While owing
through a non-magnetic material, nothing happens to both populations of electrons. STT-MRAM does
not use any magnetic field, the write selectivity is very precise (since the spin-torque effect happens
within a cell without influencing surrounding ones), thus magnetic interferences are not present
anymore. Write current is also much lower than the one required by Field MRAM as well as the
architecture is simpler (no longer multiple wires to discriminate between read and write operations).

 Physical Unclonable Function:

PUF stands for Physical Unclonable Function and identies a well-known physical characteristic which
is associated univocally to a specific structure like an integrated circuit at the physical level. The word
physical is very important, since the idea behind PUF is to rely on specific parameters at the materials
level constituting a given device to generate some useful information for identification purposes. A
PUF is a one-way function that, evaluated for each chip of a production lot, considers process
variation to generate a unique signature for that chip and that chip only, solely based on the process
variation itself. In particular,
Process variation is then the random (thus unpredictable) input x to the PUF;
The PUF is a function that is applied to the input: f(x);
The generated output, y=f(x) is the unique signature of that given chip and hopefully unique.
This response is then used for security- or identification-related goals. No more than one chip can
come with the same signature. The advantage of PUF is that it is a one-way function: this means that,
starting from the input domain, it is possible to get a univocal result y=f(x), however, the image
(result) to which y belongs, is very difficult to invert. Given y, it is very hard to find x.

 Pretty Good Privacy:

PGP stands for Pretty Good Privacy invented by Phil Zimmermann. PGP was designed to provide all
four aspects of security, i.e., privacy, integrity, authentication, and non-repudiation. It uses a digital
signature (a combination of hashing and public key encryption) to provide integrity, authentication,
and non-repudiation. PGP uses a combination of secret key encryption and public key encryption to
provide privacy. Therefore, we can say that the digital signature uses one hash function, one secret
key, and two private-public key pairs.
These are the steps to be followed to apply PGP policy for the data at rest.
 Data that needs to be encrypted
 A method to encrypt the data using a cryptographic algorithm
 Encryption keys to be used in conjunction with the data and the algorithm

3
The commonly method for encrypting and decrypting data is PGP (Pretty Good Privacy). PGP uses
symmetric and asymmetric keys to encrypt data being transferred across networks. Asymmetric
encryption uses two different keys for the encryption and decryption and both keys are derived from
one another and created at the same time. These are divided into and referred to as a public and a
private key that makes up the key pair. Data is only encrypted with a public key and thus can only be
decrypted with the matching private key. PGP provides additional security that prevents anyone who
has only the public key from decrypting data which was previously encrypted with it. Another benefit
of asymmetric encryption is that it allows authentication check. This seemed to be a viable option but
with some limitations.

PGP is mainly beneficial in cases when sensitive data is exchanged between partners; essentially
when the information is shared over the network. It works fine when you work for attaining public
key cryptography. Moreover, PGP requires more computational resources that can lead to performance
issues and make the process cumbersome.
Challenges and Future Directions:
The major challenge arises when integrating the PUF to PGP algorithm. The PUF key must me made
part of Public key and private key used by the PGP algorithm and the ways the algorithm detects it.
The MRAM has to be well processed with the algorithms used to follow integrity rules which helps is
protecting the data at rest by hackers.

5) Bibliografia / References

1. International technology roadmap for semiconductors (ITRS), emerging research devices 2013
edition. Retrieved December 23, 2014, from http://www.itrs.com.
2. Prince, B.: Emerging memories: technologies and trends. Springer Science and Business
Media, 2002.
3. Gallagher, W., Parkin, S. S., Lu, Y., Bian, X., Marley, A., Roche, K., Altman, R., Rishton, S.
Jahnes, C., Shaw, T., et al.: Microstructured magnetic tunnel junctions. Journal of Applied
Physics, 81(8):3741{3746, 1997.
4. Julliere, M.: Tunneling between ferromagnetic _lms. Physics letters A, 54(3):225{226, 1975.
5. Rodary, G., Hehn, M., Dimopoulos, T., Lacour, D., Bangert, J., Ja_r_es, H., Montaigne, F. van
Dau, F. N., Petro_, F., Schuhl, A., et al.: Development of a magnetic tunnel transistor based on
a double tunnel junction. Journal of magnetism and magnetic materials, 290:1097{1099, 2005.
6. Shuto, Y., Nakane, R., Wang, W., Sukegawa, H., Yamamoto, S., Tanaka, M., Inomata, K.
Sugahara, S.: A new spin-functional metal {oxide {semiconductor _elde_ect transistor based
on magnetic tunnel junction technology: Pseudo-spin-mosfet. Applied physics express,
3(1):013003, 2010.
7. Yuasa, S., Nagahama, T., Fukushima, A., Suzuki, Y., and Ando, K.: Giant room-temperature
magnetoresistance in single-crystal fe/mgo/fe magnetic tunnel junctions. Nature materials,
3(12):868{871, 2004.

4
8. Gallagher, W. J. and Parkin, S. S.: Development of the magnetic tunnel junction mram atibm:
From _rst junctions to a 16-mb mram demonstrator chip. IBM Journal of Research and
Development, 50(1):5{23, 2006.