Securing Networks in the Programmable Data Plane Era Luciano - - PowerPoint PPT Presentation

securing networks in the programmable data plane era
SMART_READER_LITE
LIVE PREVIEW

Securing Networks in the Programmable Data Plane Era Luciano - - PowerPoint PPT Presentation

Securing Networks in the Programmable Data Plane Era Luciano Gaspary paschoal@inf.ufrgs.br Instituto de Informtica UFRGS Securing Networks in the Programmable Data Plane Era Network softwarization : the first wave P4 programs are


slide-1
SLIDE 1

Securing Networks in the Programmable Data Plane Era

Luciano Gaspary

paschoal@inf.ufrgs.br Instituto de Informática – UFRGS

slide-2
SLIDE 2

Securing Networks in the Programmable Data Plane Era

slide-3
SLIDE 3

Network softwarization: the first wave

  • P4 programs are subject to bugs

○ Nonconformity with RFCs ○ Malformed packets ○ Use of uninitialized variables

  • Existing tools are incapable of timely verifying P4 code

3

slide-4
SLIDE 4

Network softwarization: the second wave

4

slide-5
SLIDE 5
slide-6
SLIDE 6

Problems and opportunities

  • P4 programs are subject to bugs

○ Nonconformity with RFCs ○ Malformed packets ○ Use of uninitialized variables

  • Correctness and security

properties can be violated

  • Existing tools are incapable of

timely verifying P4 code

  • We have an unprecedented
  • pportunity to devise new security services
  • W. L. C. Cordeiro, J. A. Marques, L. P. Gaspary. Dat

ata a Plan ane Program ammab ability Beyond OpenFlow: Opportunities an and Chal allenges for Network an and Service Operat ations an and Man anag

  • agement. J. Netw. Syst. Manage., v. 25, n. 4, p. 784-818, 2017.
slide-7
SLIDE 7

Assert-p4

  • Efficient verification of programmable data planes
  • Use of assertions and symbolic execution
  • Capable of verifying properties in the order of seconds
  • https://github.com/gnmartins/assert-p4
  • M. Neves, A. Schaeffer-Filho, M. Barcellos. Verificat

ation of P4 program ams in feas asible time using as

  • assertions. ACM CoNEXT 2018.
slide-8
SLIDE 8

P4box

  • P4 program monitor (guarantees properties at runtime)
  • Useful for cases where verification is impracticable
  • Instrumentation of

P4 programs during compilation

  • Low networking

device overhead

  • https://github.com/mcnevesinf/p4box
  • M. Neves, B. Huffaker, K. Levchenko, M. Barcellos. Dynam

amic pr prope perty enf enforcem cement ent in in program ammab able dat ata a plan

  • anes. IFIP NETWORKING
  • 2019. (To appear) (3r

3rd d pr prize of

  • f the

the AC ACM SIGCOMM st student resear arch co competi etiti tion 2017 2017)

slide-9
SLIDE 9

Offloading anomaly detection to P4

Packet Processing Programming Language

  • Protocol independent
  • Target independent
  • Field reconfigurable

In-network monitoring program Fine-grained measurements Real-time inspection

Challenges: line rate execution (programmable hardware switch)

Time budget: ~ dozens of nanoseconds per packet Memory space: ~50 MB SRAM, ~ 5 MB TCAM Limited programming primitives Elementary arithmetic Table lookups

How to overcome such challenges to reap the benefits of an in-network, programmable design?

  • A. Lapolli, J. A. Marques, L. P. Gaspary. Offload

ading real al-time DDoS at attac ack detection to program ammab able dat ata a plan

  • anes. IFIP/IEEE IM 2019.

(Best student pap aper aw awar ard)

slide-10
SLIDE 10
  • Entropy estimation over observation windows
  • Real-time traffic characterization based on the entropy values of the

legitimate traffic

  • In-network anomaly detection
  • https://github.com/aclapolli/ddosd-p4

Offloading anomaly detection to P4

  • A. Lapolli, J. A. Marques, L. P. Gaspary. Offload

ading real al-time DDoS at attac ack detection to program ammab able dat ata a plan

  • anes. IFIP/IEEE IM 2019.

(Best student pap aper aw awar ard)

slide-11
SLIDE 11

Ongoing/future work

11

?

  • Offloading traffic filters to

programmable switches for a more efficient strategy to triage the packets submitted to Zeek (Bro)

  • Proposal of more sophisticated

reasoning mechanisms (ML-based) for intrusion detection

  • Proposal of attack mitigation

mechanisms

slide-12
SLIDE 12

Thank you ;-)

Luciano Gaspary

paschoal@inf.ufrgs.br Instituto de Informática – UFRGS