Changeset 13 for trunk/lib


Ignore:
Timestamp:
07/19/15 15:05:42 (9 years ago)
Author:
tim
Message:

.

Location:
trunk/lib/nanownlib
Files:
3 edited

Legend:

Unmodified
Added
Removed
  • trunk/lib/nanownlib/__init__.py

    r11 r13  
    176176    my_ip = getLocalIP(target_ip, target_port)
    177177    my_iface = getIfaceForIP(my_ip)
    178     return subprocess.Popen(['chrt', '-r', '99', './bin/csamp', my_iface, my_ip,
     178    return subprocess.Popen(['chrt', '-r', '99', 'nanown-csamp', my_iface, my_ip,
    179179                             target_ip, "%d" % target_port, output_file, '0'])
    180180
     
    256256        suspect += 'R' # reordered received packets
    257257   
    258     packet_rtt = last_rcvd['observed'] - last_sent['observed']
    259     if packet_rtt < 0:
    260         sys.stderr.write("WARN: Negative packet_rtt. last_rcvd=%s,last_sent=%s\n" % (last_rcvd, last_sent))
    261 
    262258    last_sent_ack = None
    263259    try:
    264         last_sent_ack = min(((p['observed'],p) for p in packets
    265                              if p['sent']==0 and p['payload_len']+last_sent['tcpseq']==p['tcpack']))[1]
     260        last_sent_ack = min(((p['tcpack'],p['observed'],p) for p in packets
     261                             if p['sent']==0 and p['payload_len']+last_sent['tcpseq']>=p['tcpack']))[2]
    266262       
    267263    except Exception as e:
    268264        sys.stderr.write("WARN: Could not find last_sent_ack.\n")
    269265
     266    packet_rtt = last_rcvd['observed'] - last_sent['observed']
    270267    tsval_rtt = None
    271268    if None not in (timestamp_precision, last_sent_ack):
    272269        tsval_rtt = int(round((last_rcvd['tsval'] - last_sent_ack['tsval'])*timestamp_precision))
    273270
     271    if packet_rtt < 0 or (tsval_rtt != None and tsval_rtt < 0):
     272        #sys.stderr.write("WARN: Negative packet or tsval RTT. last_rcvd=%s,last_sent=%s\n" % (last_rcvd, last_sent))
     273        suspect += 'N'
     274       
    274275    return {'packet_rtt':packet_rtt,
    275276            'tsval_rtt':tsval_rtt,
     
    279280
    280281
    281 # trimean and mad for each dist of differences
     282# septasummary and mad for each dist of differences
    282283def evaluateTrim(db, unusual_case, strim, rtrim):
    283284    cursor = db.conn.cursor()
     
    292293      FROM (SELECT probes.sample s,packet_rtt FROM probes,trim_analysis WHERE sent_trimmed=:strim AND rcvd_trimmed=:rtrim AND trim_analysis.probe_id=probes.id AND probes.test_case=:unusual_case AND probes.type in ('train','test')) u
    293294    """
    294 
     295    #TODO: check for "N" in suspect field and return a flag
     296   
    295297    params = {"strim":strim,"rtrim":rtrim,"unusual_case":unusual_case}
    296298    cursor.execute(query, params)
    297299    differences = [row[0] for row in cursor]
    298300   
    299     return ubersummary(differences),mad(differences)
     301    return septasummary(differences),mad(differences)
    300302
    301303
     
    362364    for strim in range(0,num_sent):
    363365        for rtrim in range(0,num_rcvd):
     366            #print(strim,rtrim)
    364367            if strim == 0 and rtrim == 0:
    365368                continue # no point in doing 0,0 again
  • trunk/lib/nanownlib/stats.py

    r11 r13  
    166166    #return statistics.mean((l1,l2,l3,m,r3,r2,r1))
    167167
     168   
     169def septasummary(values, distance=25):
     170    left2 = 50-distance
     171    left3 = 50-(distance/2.0)
     172    left1 = left2/2.0
     173    right2 = 50+distance
     174    right3 = 50+(distance/2.0)
     175    right1 = (right2+100)/2.0
     176    l1,l2,l3,m,r3,r2,r1 = numpy.percentile(values, (left1,left2,left3,50,right3,right2,right1))
     177    return (l1+l2+l3+m+r3+r2+r1)/7.0
     178
    168179
    169180def tsvalwmean(subseries):
     
    254265ubersummaryTest = functools.partial(summaryTest, ubersummary)
    255266quadsummaryTest = functools.partial(summaryTest, quadsummary)
     267septasummaryTest = functools.partial(summaryTest, septasummary)
    256268
    257269def rmse(expected, measurements):
     
    327339        else:
    328340            return 0
     341
     342
     343from pykalman import KalmanFilter
     344def pyKalman4DTest(params, greater, samples):
     345    kp = params['kparams']
     346    #kp['initial_state_mean']=[quadsummary([s['unusual_packet'] for s in samples]),
     347    #                          quadsummary([s['other_packet'] for s in samples]),
     348    #                          numpy.mean([s['unusual_tsval'] for s in samples]),
     349    #                          numpy.mean([s['other_tsval'] for s in samples])]
     350    kf = KalmanFilter(n_dim_obs=4, n_dim_state=4, **kp)
     351    smooth,covariance = kf.smooth([(s['unusual_packet'],s['other_packet'],s['unusual_tsval'],s['other_tsval'])
     352                                   for s in samples])
     353    m = numpy.mean(smooth)
     354    if greater:
     355        if m > params['threshold']:
     356            return 1
     357        else:
     358            return 0
     359    else:
     360        if m < params['threshold']:
     361            return 1
     362        else:
     363            return 0
     364   
  • trunk/lib/nanownlib/train.py

    r11 r13  
    300300
    301301
     302from pykalman import KalmanFilter
     303_pykalman4d_params = None
     304_pykalman4d_params = {'observation_covariance': [[11960180434.411114, 4760272534.795976, 8797551081.431936, 6908794128.927051], [4760272534.795962, 12383598172.428213, 5470747537.2599745, 11252625555.297853], [8797551081.431955, 5470747537.2601185, 1466222848395.7058, 72565713883.12643], [6908794128.927095, 11252625555.297981, 72565713883.12654, 1519760903943.507]], 'transition_offsets': [592.5708159274, 583.3804671015271, 414.4187239098291, 562.166786712371], 'observation_offsets': [165.2279084503762, 157.76807691937614, 168.4235495099334, 225.33433430227353], 'initial_state_covariance': [[33599047.5, -18251285.25, 3242535690.59375, -8560730487.84375], [-18251285.25, 9914252.3125, -1761372688.59375, 4650260880.1875], [3242535690.59375, -1761372688.59375, 312926663745.03125, -826168494791.7188], [-8560730487.84375, 4650260880.1875, -826168494791.7188, 2181195982530.4688]], 'initial_state_mean': [12939012.5625, 12934563.71875, 13134751.608, 13138990.9985], 'transition_covariance': [[2515479496.145993, -401423541.70620924, 1409951418.1627903, 255932902.74454522], [-401423541.706214, 2744353887.676857, 1162316.2019491254, 1857251491.3987627], [1409951418.1628358, 1162316.2020361447, 543279068599.8229, -39399311190.5746], [255932902.74459982, 1857251491.398838, -39399311190.574585, 537826124257.5266]], 'observation_matrices': [[1.4255288693095167, -0.4254638445329988, 0.0003406844036817347, -0.0005475021956726778], [-0.46467270827589857, 1.4654311778340343, -0.0003321330280128265, -0.0002853945703691352], [-0.2644570970067974, -0.33955835481495455, 1.7494161615202275, -0.15394117603733548], [-0.3419097544041847, -0.23992883666045373, -0.15587790880447727, 1.7292393175137022]], 'transition_matrices': [[0.52163952865412, 0.47872618354122665, -0.0004322286766109684, 0.00017293351811531466], [0.5167436693545113, 0.48319044922845933, 7.765428142114672e-05, -0.00021518950285326355], [0.2091705950622469, 0.41051399729482796, 0.19341113299389256, 0.19562916616052917], [0.368592004009912, 0.22263632461118732, 0.20756792378812872, 0.20977025833570906]]}
     305_pykalman4d_good_threshold = 2009.25853272
     306_pykalman4d_params = None
     307
     308_pykalman4d_params = {'observation_covariance': [[32932883342.63772, 18054300398.442295, 27538911550.824535, 17152378956.778696], [18054300398.446983, 436546443436.5115, 37327644533.69647, 424485386677.31274], [27538911550.838238, 37327644533.706024, 3276324705772.982, 456017515263.88715], [17152378956.788027, 424485386677.317, 456017515263.88245, 3767844180658.1724]], 'observation_matrices': [[1.025773112769464, -0.028755990114063934, 0.0003540921897382532, 0.0025748564713126143], [-0.8595457826320256, 1.8607522167556567, -0.003520779053701517, 0.002309145982167138], [-0.5806427858959466, 0.22466075141448982, 1.6247192012813798, -0.27363797512617793], [-0.5853369461874607, 0.262177909212312, -0.28415108658843735, 1.6020343138710018]], 'initial_state_mean': [0.0, 0.0, 0.0, 0.0], 'observation_offsets': [549.4498515668686, 484.2106453284049, 648.556719142234, 380.10978090584763], 'transition_covariance': [[4147844406.7768326, -1308763245.5992138, 2920744388.523955, 860096280.797968], [-1308763245.5998695, 171190325905.83395, 3557618712.218984, 165332873663.83142], [2920744388.532502, 3557618712.2283373, 1054894349089.0673, -117551209299.73402], [860096280.805706, 165332873663.83963, -117551209299.73474, 1223605046475.7324]], 'transition_offsets': [1156.9264087977374, 1150.752680207601, 1312.2595286459816, 1267.4069537452415], 'initial_state_covariance': [[667999273207241.0, 669330484615232.1, 713726904326576.2, 731731206363217.4], [669330484615390.9, 670664348906228.8, 715149243295271.9, 733189424910272.2], [713726904326843.4, 715149243295370.6, 762584802695960.9, 781821582244358.5], [731731206363417.0, 733189424910299.0, 781821582244278.6, 801543624134758.0]], 'transition_matrices': [[0.9680677036616316, 0.03260717171917804, 0.0005279411071512641, -0.0012363486571871363], [0.9555219601128613, 0.03851351491891819, 0.00411268796118236, 0.0017357967358293536], [0.622254432930994, -0.2583795512595657, 0.31745705251401546, 0.32357126976364725], [0.6644076824932768, -0.33545285094373867, 0.3295778964272671, 0.34682391469482354]]}
     309_pykalman4d_good_threshold = -253.849393803
     310def trainPyKalman4D(db, unusual_case, greater, num_observations):
     311    global _pykalman4d_params
     312    global _pykalman4d_good_threshold
     313    db.resetOffsets()
     314
     315    if _pykalman4d_params == None:
     316        train = db.subseries('train',unusual_case, offset=0)
     317        null = db.subseries('train_null',unusual_case, offset=0)
     318        train_array = numpy.asarray([(s['unusual_packet'],s['other_packet'],s['unusual_tsval'],s['other_tsval'])
     319                                     for s in train])
     320        null_array = numpy.asarray([(s['unusual_packet'],s['other_packet'],s['unusual_tsval'],s['other_tsval'])
     321                                    for s in null])
     322        kf = KalmanFilter(n_dim_obs=4, n_dim_state=4)
     323        #initial_state_mean=[quadsummary([s['unusual_packet'] for s in train]),
     324        #                                      quadsummary([s['other_packet'] for s in train]),
     325        #                                      numpy.mean([s['unusual_tsval'] for s in train]),
     326        #                                      numpy.mean([s['other_tsval'] for s in train])])
     327
     328        kf = kf.em(train_array+null_array[0:50000], n_iter=10,
     329                   em_vars=('transition_matrices',
     330                            'observation_matrices',
     331                            'transition_offsets',
     332                            'observation_offsets',
     333                            'transition_covariance',
     334                            'observation_covariance',
     335                            'initial_state_covariance'))
     336        _pykalman4d_params = {'transition_matrices': kf.transition_matrices.tolist(),
     337                              'observation_matrices': kf.observation_matrices.tolist(),
     338                              'transition_offsets': kf.transition_offsets.tolist(),
     339                              'observation_offsets': kf.observation_offsets.tolist(),
     340                              'transition_covariance': kf.transition_covariance.tolist(),
     341                              'observation_covariance': kf.observation_covariance.tolist(),
     342                              'initial_state_mean': kf.initial_state_mean.tolist(),
     343                              'initial_state_covariance': kf.initial_state_covariance.tolist()}
     344        print(_pykalman4d_params)
     345   
     346        kf = KalmanFilter(n_dim_obs=4, n_dim_state=4, **_pykalman4d_params)
     347        smoothed,covariance = kf.smooth(train_array)
     348        null_smoothed,covariance = kf.smooth(null_array)
     349
     350        kp = _pykalman4d_params.copy()
     351        #kp['initial_state_mean']=[quadsummary([s['unusual_packet'] for s in train]),
     352        #                          quadsummary([s['other_packet'] for s in train]),
     353        #                          numpy.mean([s['unusual_tsval'] for s in train]),
     354        #                          numpy.mean([s['other_tsval'] for s in train])]
     355        #kf = KalmanFilter(n_dim_obs=4, n_dim_state=4, **kp)
     356        #null_smoothed,covariance = kf.smooth(null_array)
     357       
     358        _pykalman4d_good_threshold = (numpy.mean([m[0]-m[1] for m in smoothed])+numpy.mean([m[0]-m[1] for m in null_smoothed]))/2.0
     359        print(_pykalman4d_good_threshold)
     360
     361   
     362    def trainAux(params, num_trials):
     363        estimator = functools.partial(pyKalman4DTest, params, greater)
     364        estimates = bootstrap3(estimator, db, 'train', unusual_case, num_observations, num_trials)
     365        null_estimates = bootstrap3(estimator, db, 'train_null', unusual_case, num_observations, num_trials)
     366       
     367        bad_estimates = len([e for e in estimates if e != 1])
     368        bad_null_estimates = len([e for e in null_estimates if e != 0])
     369       
     370        false_negatives = 100.0*bad_estimates/num_trials
     371        false_positives = 100.0*bad_null_estimates/num_trials
     372        return false_positives,false_negatives
     373
     374    params = {'threshold':_pykalman4d_good_threshold, 'kparams':_pykalman4d_params}
     375
     376    wt = WorkerThreads(2, trainAux)
     377    num_trials = 50
     378    performance = []
     379    for t in range(-80,100,20):
     380        thresh = _pykalman4d_good_threshold + abs(_pykalman4d_good_threshold)*(t/100.0)
     381        params['threshold'] = thresh
     382        wt.addJob(thresh, (params.copy(),num_trials))
     383    wt.wait()
     384    while not wt.resultq.empty():
     385        job_id,errors = wt.resultq.get()
     386        fp,fn = errors
     387        #performance.append(((fp+fn)/2.0, job_id, fn, fp))
     388        performance.append((abs(fp-fn), job_id, fn, fp))
     389    performance.sort()
     390    #pprint.pprint(performance)
     391    best_threshold = performance[0][1]
     392    #print("best_threshold:", best_threshold)
     393    params['threshold']=best_threshold
     394
     395    wt.stop()
     396   
     397    return {'trial_type':"train",
     398            'num_observations':num_observations,
     399            'num_trials':num_trials,
     400            'params':json.dumps(params, sort_keys=True),
     401            'false_positives':performance[0][3],
     402            'false_negatives':performance[0][2]}
     403
     404
     405
    302406classifiers = {'boxtest':{'train':trainBoxTest, 'test':multiBoxTest, 'train_results':[]},
    303407               'midsummary':{'train':functools.partial(trainSummary, midsummary), 'test':midsummaryTest, 'train_results':[]},
    304                'ubersummary':{'train':functools.partial(trainSummary, ubersummary), 'test':ubersummaryTest, 'train_results':[]},
     408               #'ubersummary':{'train':functools.partial(trainSummary, ubersummary), 'test':ubersummaryTest, 'train_results':[]},
    305409               'quadsummary':{'train':functools.partial(trainSummary, quadsummary), 'test':quadsummaryTest, 'train_results':[]},
    306                'tsvalwmean':{'train':trainTsval, 'test':tsvalwmeanTest, 'train_results':[]},
     410               'septasummary':{'train':functools.partial(trainSummary, septasummary), 'test':septasummaryTest, 'train_results':[]},
     411               #'pykalman4d':{'train':trainPyKalman4D, 'test':pyKalman4DTest, 'train_results':[]},
     412               #'tsvalwmean':{'train':trainTsval, 'test':tsvalwmeanTest, 'train_results':[]},
    307413               #'kalman':{'train':trainKalman, 'test':kalmanTest, 'train_results':[]},
    308414               #'_trimean':{'train':None, 'test':trimeanTest, 'train_results':[]},
Note: See TracChangeset for help on using the changeset viewer.