def a3(traces, referenceTrace, depth=-1):
from tools import sequence, algorithms
seqRef = sequence.Sequence(algorithms.removeRepetitions(referenceTrace.getNormalizedSequence(depth)))
seqRef.removeAll(referenceTrace.getVirtualStartingNode())
seqs = []
for trace in traces:
seq = sequence.Sequence(algorithms.removeRepetitions(trace.getNormalizedSequence(depth)))
seq.removeAll(trace.getVirtualStartingNode())
seqs.append(seq)
return DecisionNodeFinding._a3core(seqs, seqRef)
def a0(traces, referenceTrace):
"""
Assumes the trace for the invalid command as fixed-point. Consists of the following steps:
1) Normalizes all traces and removes repetitions.
2) Determines the largest common sub-sequence of the trace of the invalid command and each valid command.
3) Determines the largest common sub-sequence of all previously determined invalid command/command sub-sequences
4) Outputs the sequence determined in step 3) in reverse order. The first function in this output is likely to be the command-dispatcher. If not, it could likely be the functions at the subsequent positions.
Functions should get a score depending on their position in the sequence (lower == better) and the overall lenght of the sequence (longer == better).
If the returned list is empty, it is likely that the server-application in question does not possess a central command-dispatcher (maybe multiple for different commands or chained command-handlers with each having its own dispatcher).
Possible further improvements:
1) If the length of the lcss drops significantly for a certain subseq, then this subseq should be left out, at the cost of a lower scoring for the result.
@param traces: A list of function traces belonging to a certain protocol string
@type traces: pygdb.trace.FunctionTrace
@param referenceTrace: A trace
"""
from tools import sequence, algorithms
seqRef = sequence.Sequence(algorithms.removeRepetitions(referenceTrace.getNormalizedSequence()))
seqRef.removeAll(referenceTrace.getVirtualStartingNode())
subSeqsInvalid = []
for trace in traces:
if trace is None:
continue
seq = sequence.Sequence(algorithms.removeRepetitions(trace.getNormalizedSequence()))
seq.removeAll(trace.getVirtualStartingNode())
lcss = seqRef.findLargestCommonSubSequence(seq)
if lcss is not None:
subSeqsInvalid.append(lcss)
if len(subSeqsInvalid) == 0:
return []
lcss = subSeqsInvalid[0]
for subSeqInvalid in subSeqsInvalid[1:]:
if lcss is None:
return []
lcss = lcss.findLargestCommonSubSequence(subSeqInvalid)
if lcss is None:
return []
return _Helper.normalizeScores({lcss.rawSeq[i] : i for i in range(len(lcss.rawSeq))})
