19
Informa(on Assurance in a Distributed Forensic Cluster Nick Pringle a *, Mikhaila Burgess a a University of South Wales (formerly University of Glamorgan), Treforest, CF37 1DL, UK

Informaon)Assurance) in)a Distributed)Forensic) Cluster)old.dfrws.org/.../presentations/DFRWS2014-pres3.pdf · 2014-08-09 · Informaon)Assurance) in)a Distributed)Forensic) Cluster))

  • Upload
    others

  • View
    3

  • Download
    0

Embed Size (px)

Citation preview

Informa(on  Assurance  in  a  

Distributed  Forensic  Cluster  

 

•  Nick  Pringlea*,  Mikhaila  Burgessa  

•  a  University  of  South  Wales  (formerly  University  of  Glamorgan),  Treforest,  CF37  1DL,  UK  

•  This  is  a  short  presenta(on  of  the  work  presented  at  DFRWS  Europe  2014  

•  www.fcluster.org.uk  •  PhD  published  at  the  end  of  the  year  

ComplexHighly SkilledSmall Quantity

SimpleBasic Skills

Large Quantity

Opera(on  Big  Wing,  24th  April  2014  3,300  Metropolitan  Police  Officers  Targe(ng  Co-­‐ordinated  arrests  of  630  persons  across  London  as  burglary  and  theS  crackdown    Na(onal  Crime  Agency  Opera(on  Notorise  resulted  in  660  persons  arrested  in  regards  child  abuse.    Specifically  9,172  devices  seized.  

Source  media  Write  

Protected  Copy  

Image  on  Local  

harddisk  

Artefact  extrac(on  Processing  

Analysis  Repor(ng  

8GB  USB  2  S(ck  

SATA  III  but  HD  85MB/s  

i7  12  cores  

Analysis  Repor(ng  

2  MBytes/sec  per  core  20  minutes  

3TB  hard  Disk  75  Mbyte/s  

SATA  III  but  HD  85MB/s  

i7  12  cores  

Analysis  Repor(ng  

3TB  hard  Disk  75  Mbyte/s  

SATA  III  SSD  at  450  MB/s  

i7  12  cores  

Analysis  Repor(ng  

$$$$  Expensive  

Results  Database  

Results  Database  

Results  Database  

Results  Database  

Our    Media  

Our    Media  

Write  Protected  Copy  

Direct  Write  Protected  Copy  

Direct  Write  Protected  Copy  

Copy  

Copy  

11  hours  

2  minutes  

3TB  hard  Disk  75  Mbyte/s  

SATA  III  but  HD  85MB/s  

i7  12  cores  

Analysis  Repor(ng  

Results  Database  

Our    Media  

Copy  Write  Protected  Copy  

11  hours  2  MBytes/sec  per  core  70  hours  

S7ll  2  MBytes/sec  per  core  S7ll  70  hours  

2  minutes  

11  hours  

S7ll  2  MBytes/sec  per  core  S7ll  70  hours  

Processor  Bound  

11  hours  

No  Control  over  Source  device  speed  All  Cores  at  100%  

Chain  of  Evidence  Wait  

Wait   Wait   Wait  

3TB  hard  Disk  75  Mbyte/s  

SATA  III  SSD  at  450  MB/s  

i7  12  cores  

11  hours   2  MBytes/sec  per  core  25  x  i7  =  300  cores  2  hours    

Gigabit  Network  100  MB/s  =  10  hours  

3TB  hard  Disk  75  Mbyte/s  

SATA  III  SSD  at  450  MB/s  

Xeon  64  cores  

11  hours   2  MBytes/sec  per  core  13  hours  

$$$$  Expensive!  $150,000?  Unacceptable  SGI  Al(x  4700?  

i7  12  cores  i7  

12  cores  i7  12  cores  i7  

12  cores  10  Gigabit  Network  1000  MB/s  $$$  Expensive  For  25  W/S  $25,000  

Analysis  Repor(ng  

Analysis  Repor(ng  

Results  Database  

Results  Database  

Direct  Write  Protected  Copy  

Direct  Write  Protected  Copy  

i7  12  cores  i7  12  cores  i7  

12  cores  i7  12  cores  i7  

12  cores  

This  problem  is  not  going  away.  It’s  going  to  get  worse!  RCFL  Figures  

2004   2012   %   2020?  

Examinations 1304   8566   657   56270  Total Volume Examined

229TB   5886  TB   2570   151  EB  Average Case 175GB   680  GB   300   2.6  TB  

Linear  Processing  

Mul(-­‐Threading/  Parallel/  Distributed  Processing  

Task  setup   Processing   Task  closure  

Latency,  Mul(-­‐threading    and  Parallel  Processing  

I7  -­‐  12  cores  Hard  Disk  

Source  media  

Write  Protect  

Artefact  extrac(on  Processing  

Analysis  Repor(ng  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

Source  media  

Write  Protect  

Source  media  

Write  Protect  

Source  media  

Write  Protect  

Results  Database  

I7  -­‐  12  cores  Hard  Disk  

Source  media  

Write  Protect  

Artefact  extrac(on  Processing  

Analysis  Repor(ng  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

Source  media  

Write  Protect  

Source  media  

Write  Protect  

Source  media  

Write  Protect  

Results  Database  

DEB   DEB   DEB   DEB   DEB   DEB   DEB  DEB  

DEB  

DEB   DEB   DEB   DEB   DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  DEB  

DEB  DEB  

DEB   DEB  

DEB  DEB  

DEB  DEB  

DEB  DEB  DEB  

DEB  

DEB   DEB   DEB   DEB   DEB  

DEB   DEB   DEB   DEB   DEB  

We  lose  “Chain  of  Evidence”  •  We’re  not  longer  using  a  simple  system    

   with  one  file  store  and  a  few  PCs  

•  In  this  world  of  distributed  storage  and  processing      we  need  to  revisit  and  re-­‐establish      “Chain  of  Evidence”  within  the  computer  system  

•  We’re  back  a  decade  and  can’t  move  on  un(l  we  do  

I7  -­‐  12  cores  Hard  Disk  

Source  media  

Write  Protect  

Artefact  extrac(on  Processing  

Analysis  Repor(ng  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

Source  media  

Write  Protect  

Source  media  

Write  Protect  

Source  media  

Write  Protect  

Results  Database  (Hadoop?)  

DEB   DEB   DEB   DEB   DEB   DEB   DEB  DEB  

DEB  

DEB   DEB   DEB   DEB   DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  DEB  

DEB  DEB  

DEB   DEB  

DEB  DEB  

DEB  DEB  

DEB  DEB  DEB  

DEB  

DEB   DEB   DEB   DEB   DEB  

DEB   DEB   DEB   DEB   DEB  “Jigsaw”  Im

aging  

Data  Acquisi7o

n  Triage  

FClusterfs  

I7  -­‐  12  cores  Hard  Disk  

Source  media  

Artefact  extrac(on  Processing  

Analysis  Repor(ng  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

I7  -­‐  12  cores  Hard  Disk  

Artefact  extrac(on  Processing  

Source  media  

Source  media  

Source  media  

Results  Database  (Hadoop?)  

DEB  DEB  

DEB  

DEB  DEB  

DEB  DEB  

DEB  DEB  

“Jigsaw”  Imaging  Data  Acquisi7on  Triage  

FClusterfs  

Processing  Server  Results  Meta  Data  table  

Replicator  Data  Storage  Server  

FClusterfs  metadata  storage  Metadata  Import/Load  Balancer  

Acquisi(on  Authority  Jigsaw  Imager  and  Data  Triage  

FCluster  Architecture  Roles  and  Zones  

Jigsaw  Imaging  •  Reads  the  Source  media  and  follows  the  file  system  (not  sector  by  sector)  

•  Creates  Digital  Evidence  Bags  for  each  file  while  simultaneously  crea(ng  the  conven(onal  image  

Source  media  

Write  Protect  

Source  media  

Write  Protect  

Source  media  

Write  Protect  

Source  media  

Write  Protect  

DEB   DEB   DEB   DEB   DEB   DEB   DEB  DEB  

DEB  

DEB   DEB   DEB   DEB   DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  DEB  

DEB  DEB  

DEB   DEB  

DEB  DEB  

DEB  DEB  

DEB  DEB  DEB  

DEB  

DEB   DEB   DEB   DEB   DEB  

DEB   DEB   DEB   DEB   DEB  

Data  Acquisi(on  Triage  

•  Uses  a  Bayesian  approach  that  directs  the  Jigsaw  Imaging  process  to  priori(se  files  considered  to  be  of  higher  likelihood  of  yielding  evidence  

Source  media  

Write  Protect  

Source  media  

Write  Protect  

Source  media  

Write  Protect  

Source  media  

Write  Protect  

DEB   DEB   DEB   DEB   DEB   DEB   DEB  DEB  

DEB  

DEB   DEB   DEB   DEB   DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  

DEB  DEB  

DEB  DEB  

DEB   DEB  

DEB  DEB  

DEB  DEB  

DEB  DEB  DEB  

DEB  

DEB   DEB   DEB   DEB   DEB  

DEB   DEB   DEB   DEB   DEB  

Fclusterfs  A  file  system  for  Digital  Evidence  Bags  

A  FUSE  file  system  that:  •  Stores  the  original  file  meta-­‐data  in  the  file  system  •  Gives  access  to  files  stored  as  whole,  encrypted  DEBs  •  Has  access  control  by  user  and  file  system  and  file  •  Is  Read  Only  •  Logs  movement  of,  and  access  to,  data  •  Allows  (most)  unaltered  legacy  soSware  •  Allows  non-­‐parallel-­‐aware  soSware  to  run  across  mul(ple  nodes  

Why  is  this  the  right  approach?  •  This  could  be  achieved  within  an  applica(on  program  but  

each  package  would  to  have  to  implement  it  and  gain  approval.  

•  Working  at  file  system  level  the  efficacy  is  global  •  Interac(on  with  FClusterfs  is  unavoidable  •  Fclusterfs  controls  data  access  and  maintains  Assurance  

Funded  by…  

Informa(on  Assurance  in  a  

Distributed  Forensic  Cluster  Ques(ons?