recovering data f rom corrupt packets

Recovering Data From Corrupt PacketsSENSYS 2013EWSN 2014

Typical procedure of transmissionTechniques for overcoming bit errors

ARQ : Automatic Repeat reQuestFEC : Forward Error CorrectionOr the combination : HARQ

Interesting things happened when it finds a corrupted packet during the CRC detection process!All IS NOT LOST(ANL): a opportunistic approachTVA: an novel method of CRC correcting

DSSS in 802.15.4 StandardA byte = 2 four-bit symbols

Echo symbol is finally represented by a 32-bit log chip sequence (i.e. the rightmost table)

What happened when takes noise into consideration ?Corruption may happen

ANL : ObservationDeployment

16 nodesOne year from 2012 to 2013

Mutation Matrix:Each entry denotes a mutation frequency

One Observation:Mutations are not uniformly distributed

ANL : Recovering DataConsider a received symbol

E.g s13,p(s5|s13)=

Consider a received sequence of symbolsE.g receive symbol sequence r = (s13,s3,s0,s11) a possible sent sequence t = (s5,s3,s1,s11)

Key idea: compute a probability distribution over the possible sent packets

ANL : Left questionsHow to generate the mutation matrix?

Through Simulation

How to estimate pchip : using LQI

Is the actual sent word is assigned the highest rank?(95% of them are)

TVA-A reliable protocol Procedure

Feature:A novel method of CRC error correctionCompact and computationally efficientDesigned to correct the most common error

patterns observed in WSNs

CRC error correction TheoryTabular Method

m2 - the received msg ; m1 – the sent msg ; e – the error pattern we want to correct

Prebuild a table using CRCs of target error pattern P as indexIndex into the table with c and get error sequence e.The corrected message

Cyclic Method anyway,reduced memory requirement needed by single Tabular Method e.g 64 bytes to correct 4-bit-bursts in 40 bytes of data , while the tabular method uses 10KB

…… CRC(e1) CRC(e2) CRC(e3) ……

...... e1 e2 e3 ……

Reliability of TVA:verifyCorrection process:m1

Risk : if CRC(e1) = CRC(e2) and

Verify : compare the sender side CRC and receiver side CRC

So what is this polynomial P?

Not the correcting polynomial G : since CRC(m1 e1 e2 ) = CRC(m1) CRC(e1) CRC(e2) = CRC(m1)

The key property of the polynomial P : it does not fail to detect a message error consisting of a common channel error XOR’d with a correction error

⊕e 1 e2…… CRC(e2)=c ……

…… e2 ……

Searching for Polynomials:results

Performance Comparison

Conclusion of two workSolve the same issue

Recovering information from corrupt packetDifferent method

ANL : a probabilistic approach TVA : CRC error correcting

Combination Possibility?

谢谢！

Comparison with other protocol

Outline1.All is not lost

the distinct pattern of how 802.15.4 packets are corrupted

an probabilistically approach infers the original content of a corrupt packet

2.Unlocking the full power of the CRC

a novel method of CRC error correction

a protocol TVA making use of the error correcting ability of CRC

Comparision with ANL

ObservationsObservation 1:Mutations are not uniformaly distributed

Observation 2:The most significant bit of a symbol is more stable than other bits

Observation 3:Symbols s0 to s7 are more stable than other symbols

Explanation

ExploitationCOMPUTE A DISTRIBUTION: