18
ERASE- EntRopy-based SAnitization of Sensitive Data for Privacy Preservation Presenter: Dr. Avinash Srinivasan Authors: Jeffrey Medsger and Avinash Srinivasan

Erase icitst

Embed Size (px)

DESCRIPTION

 

Citation preview

Page 1: Erase icitst

ERASE- EntRopy-based SAnitization of Sensitive Data for Privacy Preservation

Presenter: Dr. Avinash SrinivasanAuthors: Jeffrey Medsger and Avinash Srinivasan

Page 2: Erase icitst

Contributions to the Field of Computer Media Sanitization

Goal:

Design a more efficient process for non-destructively sanitizing (overwriting) computer media.

ERASE - An entropy-based wiping method for sanitizing a target area.

ERASERS - Random sampling in tandem with the above entropy-based technique for efficiently sanitizing a target area.

Tool Implemented with aforementioned capabilities.

Medsger and Srinivasan - 2

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

Performance Results

Thoroughness of Technique

Conclusion

Page 3: Erase icitst

Overview of Field: Computer Media Sanitization

Deleted Files

Sanitization Destructive

Non-destructive

Data carving can recover deleted data from unallocated space and file slack.

Study performed by Garfinkel et. al. [3] With little financial expenditure, the authors were able to retrieve

thousands of credit card numbers and other personal information.

[3] S. L. Garfinkel and A. Shelat, “Remembrance of data passed: A study of disk sanitization practices,” IEEE Security and Privacy, vol. 1, pp. 17–27, 2003.

Medsger and Srinivasan - 3

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

Performance Results

Thoroughness of Technique

Conclusion

Page 4: Erase icitst

Problem Statement Current wipe methods use a brute force approach

[overwrite everything]

Hard drive sizes increasing

In 2011, Seagate reported 590GB avg. size HDD

Medsger and Srinivasan - 4

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

Performance Results

Thoroughness of Technique

Conclusion

Suppose 50% of area to be sanitized contains sensitive data [590GB HDD]:

Wipe Pattern

Current tools [dd]

14.6hrs Pseudorandom

ERASE 9.5hrs Pseudorandom

Current tools [dd]

1.584hrs Zeros

ERASERS 0.85hrs (best case) - 1.580hrs (worst case)

Zeros

Page 5: Erase icitst

New Techniques ERASE

Calculate the entropy of each disk block and if that entropy falls within a sensitive entropy range, then overwrite that disk block.

ERASERS Divide the area (population) to be sanitized into z

subpopulations. For each subpopulation, randomly sample n blocks, and if at

least one sampled block is within the sensitive entropy range, then overwrite that entire subpopulation.

Medsger and Srinivasan - 5

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

Performance Results

Thoroughness of Technique

Conclusion Area to be sanitized

Divide into subpopulationsRandomly sample n disk blocks from each subpopulation

Overwrite subpopulations found to have sensitive data

Page 6: Erase icitst

ERASE Designed to allow for a more efficient wipe in situations

where pseudorandom data is being used as the wipe pattern.

Reduce the number of writes by performing extra reads.

Perform a read (106.8MB/s) to possibly prevent a write using data from /dev/urandom (11.8MB/s).

Also, designed to allow for a more efficient wipe when multiple overwrites are performed.

Medsger and Srinivasan - 6

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

Performance Results

Thoroughness of Technique

Conclusion

Page 7: Erase icitst

ERASERS Limitation of ERASE is that if a uniform pattern is used

as the wipe pattern, such as /dev/zero, ERASE will not be more efficient when only one overwrite pass is used.

Assuming read and write speeds are similar on drive.

ERASERS was designed to allow for a more efficient wipe when a uniform wipe pattern is used.

Accomplished by randomly sampling blocks in areas on the drive and only overwriting those areas which are found to have sensitive data.

Thus, a limited number of reads are performed to possibly prevent writes.

Envision ERASERS could be used to periodically wipe unallocated space.

Medsger and Srinivasan - 7

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

Performance Results

Thoroughness of Technique

Conclusion

Page 8: Erase icitst

Background Sanitization Algorithms

Entropy – Compressibility of Data

Random Sampling

Medsger and Srinivasan - 8

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

Performance Results

Thoroughness of Technique

Conclusion

Page 9: Erase icitst

ERASE Process1. Number of wipe passes

2. Wipe pattern - e.g., /dev/zero or /dev/urandom

3. Sensitive entropy range for the target area

4. Random sampling confidence level and interval (if using ERASERS)

Medsger and Srinivasan - 9

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

Performance Results

Thoroughness of Technique

Conclusion

Page 10: Erase icitst

One Pass or Multiple Passes

Guttman [6] states overwritten data can be read by specialized equipment, such as magnetic microscopes. Has been severe criticism of this idea in the Forensic community.

Numerous other wipe pass recommendations Proposed Method significantly outperforms Brute Force Method

using NIST SP800-88 1 pass random recommendation Performance even better with multi-pass wipes

[6] P. Gutmann, “Secure deletion of data from magnetic and solid-state memory,” in Proceedings of the 6th USENIX Security Symposium, 1996, pp. 77–89.

Medsger and Srinivasan - 10

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

One Pass or Multiple Passes

Wiping – Zeros vs. Random Patterns

Sensitive entropy range

Performance Results

Thoroughness of Technique

Conclusion

Publication Wipe Method

NIST SP800-88 1 pass with random data

Original DoD 5220.22-M 3 pass wipe

New DoD 5220.2-M Doesn’t specify

Page 11: Erase icitst

Wiping – Zeros vs. Random Patterns

Medsger and Srinivasan - 11

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

One Pass or Multiple Passes

Wiping – Zeros vs. Random Patterns

Sensitive entropy range

Performance Results

Thoroughness of Technique

Conclusion

Pattern What does that tell the second owner of the drive/forensic analyst?

Zeros Drive has not been used

Drive has been wiped

Random Difficult to tell if wiped or just high entropy data [jpg, zip, pdf]

Page 12: Erase icitst

Sensitive entropy range User tunable parameter (0 – 8 bPB)

All blocks that have an entropy greater than zero. All blocks of a certain file type. All blocks that have an entropy between a very low entropy

and a very high entropy

Medsger and Srinivasan - 12

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

One Pass or Multiple Passes

Wiping – Zeros vs. Random Patterns

Sensitive entropy range

Performance Results

Thoroughness of Technique

Conclusion

Page 13: Erase icitst

Performance Results Test Setup

ERASE/EWNS Test – x% of blocks in the partition were filled with data within the sensitive entropy range. Tool ran in ERASE/EWNS mode.

ERASERS/EWSBC (Best Case) Test – x% of the blocks in the partition were sequentially filled with data within the sensitive entropy range. Tool ran in ERASERS/EWS mode with 95% CL and 5% CI. 8 subpopulations of 4GB. Best case because data clustered compactly and sequentially.

ERASERS/EWSWC (Worst Case) Test – x% of the blocks in the partition were filled with data within the sensitive entropy range. However, the blocks were spread out to equally fill the drive, causing all subpopulations to have data within the sensitive entropy range.

Medsger and Srinivasan - 13

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

Performance Results

Test Setup

ERASE /EWNS vs. DD

ERASERS/EWS vs. DD

Thoroughness of Technique

Conclusion

Page 14: Erase icitst

ERASE/EWNS vs. DD (/dev/urandom)

Medsger and Srinivasan - 14

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

Performance Results

Test Setup

ERASE /EWNS vs. DD

ERASERS/EWS vs. DD

Thoroughness of Technique

Conclusion

Page 15: Erase icitst

ERASERS/EWS vs. DD (/dev/zero)

Medsger and Srinivasan - 15

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

Performance Results

Test Setup

ERASE /EWNS vs. DD

ERASERS/EWS vs. DD

Thoroughness of Technique

Conclusion

Page 16: Erase icitst

Thoroughness of Technique Probability of not sampling a sensitive block, when 384

blocks are randomly sampled from a population of 1048576 blocks. 384 out of 1048576 == 95% CL and 5% CI Probability evaluated for scenarios when there are 10, 100,

1000, and 10000 sensitive blocks in a population of 1048576 blocks.

Medsger and Srinivasan - 16

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

Performance Results

Thoroughness of Technique

Conclusion

Page 17: Erase icitst

Conclusion Presented a novel disk sanitization technique.

ERASE calculates the entropy of data blocks in the target area to decide whether the blocks need to be overwritten.

ERASERS enhances ERASE by using random sampling to further optimize the wiping process.

Medsger and Srinivasan - 17

ICITST-2012

ERASEContributions to Field

Overview of Field

Problem Statement

New Techniques

ERASE

ERASERS

Background

ERASE Process

Performance Results

Thoroughness of Technique

Conclusion

Page 18: Erase icitst

ICITST-201218

Questions?

Thank you!!!