Reliable first-seen dates are important for constructing temporal malware evaluation splits, but this information often comes from external services that are not equally accessible to all researchers. We investigate whether machine learning can instead estimate an executable's first-seen date or determine when its PE/COFF compilation timestamp is unreliable. Using EMBER2024, we evaluate feature-based and raw-byte models across classification, regression, ranking, and latent-reliability formulations. Direct first-seen regression produces errors comparable to a substantial portion of the dataset's collection window, and none of the evaluated scores provides a useful chronological ordering of files. Current models therefore cannot replace externally sourced first-seen metadata. In contrast, classifiers can identify timestamps that fall within specified tolerances of first seen and can detect clearly implausible timestamps with high discrimination. We also find substantial timestamp disagreement among both benign and malicious executables, indicating that unreliable timestamps are not exclusively a product of malicious manipulation. These results establish baselines and identify timestamp reliability as a distinct direction for future security-ML research.
Figures & tables
Figure 1: Graphical model for the timestamp reliability model in Algorithm 1 . Distribution boxes denote the two Student- t likelihood components for Oi=Ci−Fi ; the mixture box denotes the marginalized likelihood that combines them using wi .
Figure 2: Signed lag quantiles for benign and malicious PE/.NET files. Positive values indicate that the compile timestamp predates first seen. Negative values indicate a compile timestamp after first seen.
Figure 3: Macro lag-category shares by label. The “compile after” category counts rows where the COFF timestamp is after first seen.
Threshold
Raw share
No-COFF AUC
No-COFF F1
With-COFF AUC
With-COFF F1
30d
0.2211
0.9705
0.8337
0.9756
0.8596
60d
0.2459
0.9762
0.8606
0.9734
0.8863
90d
0.2623
0.9774
0.8690
0.9743
0.9039
180d
0.2967
0.9810
0.8947
0.9849
0.9419
Table 1: Regular test split threshold classification. With-COFF models received the observed COFF timestamp as an input feature; no-COFF models did not.
Threshold
Raw share
No-COFF AUC
No-COFF F1
With-COFF AUC
With-COFF F1
30d
0.3504
0.9734
0.8085
0.9794
0.9378
60d
0.4132
0.9671
0.8270
0.9844
0.9448
90d
0.4579
0.9680
0.8634
0.9595
0.9450
180d
0.5587
0.9810
0.8967
0.9832
0.9796
Table 2: Post-train challenge threshold classification. Raw share is the fraction of post-train challenge rows whose COFF timestamp is within the threshold.
Target/loss
Representation
COFF as predictor
MAE
MedAE
p90
raw timestamp
COFF baseline
–
6,396.68
2,681.86
20,026.00
ordinal abs-delta
XGBoost
no
2,747.66
1,486.89
7,163.55
ordinal abs-delta
XGBoost
yes
2,044.67
1,147.65
7,593.38
ordinal abs-delta
MalConv2
raw bytes
3,294.74
1,548.88
9,886.17
signed delta
XGBoost
no
2,029.49
274.33
5,351.69
signed delta
XGBoost
yes
296.33
154.05
781.46
Table 3: Regular test split point-estimation errors, in days, grouped by target or loss family.
Target/loss
Representation
COFF as predictor
MAE
MedAE
p90
raw timestamp
COFF baseline
–
2,591.26
127.01
10,817.47
ordinal abs-delta
XGBoost
no
3,008.55
3,103.88
4,685.11
ordinal abs-delta
XGBoost
yes
1,453.19
1,119.91
2,169.12
ordinal abs-delta
MalConv2
raw bytes
6,953.36
1,451.44
20,000.93
signed delta
XGBoost
no
1,997.83
1,056.79
5,150.41
signed delta
XGBoost
yes
311.82
181.83
716.57
Table 4: Post-train challenge point-estimation errors, in days, grouped by target or loss family.
Split
Score family
Representation
Pair acc.
Spearman
Median rank err.
Median position-time err.
test
raw timestamp
COFF baseline
0.5346
0.1041
0.2822
23.73
test
latent age score
PyMC-BART
0.5334
0.0852
0.2789
23.44
test
within90 probability
XGBoost no-COFF
0.5287
0.0838
0.2591
21.83
test
within90 probability
XGBoost with-COFF
0.5208
0.0603
0.2739
23.10
test
ordinal abs-delta
XGBoost no-COFF
0.5109
0.0309
0.2832
23.88
test
ordinal abs-delta
XGBoost with-COFF
0.5103
0.0298
0.2803
23.63
Table 5: Temporal ordering induced by model scores, grouped by score family. Challenge rows are restricted to files first seen after the train maximum.
Figure 4: Ordering quality from model scores on the regular test split and post-train challenge subset. Lower rank error is better.
Figure 5: Chronology curves after sorting rows by selected model scores. Each point is the median true first-seen offset within a predicted-order bin. The oracle line is the corresponding true sorted order. The challenge panel is restricted to the post-train subset.
Figure 6: Posterior interpretation of the graphical timestamp model by apparent lag bucket. The left panel shows posterior truthful-timestamp probability and the share of rows with probability at least 0.5. The right panel shows the posterior median latent compile age before first seen.
Group
N
Mean Ptruth
Median Ptruth
Share Ptruth≥0.5
Median Ai
days
all
2,000
0.2641
0.0000
0.2640
2.39
not obviously bad
1,329
0.3975
0.0060
0.3973
2.47
bad rule
671
0.0000
0.0000
0.0000
2.20
benign
1,033
0.3154
0.0002
0.3175
2.78
malicious
967
0.2093
0.0000
0.2068
2.03
Table 6: Interpretable posterior summaries from the graphical timestamp model. “Bad rule” rows are forced into the modified/artifact component.
Figure 7: Absolute first-seen prediction absolute-error quantiles by true first-seen month. The dashed line marks the month containing the training maximum first-seen date, 2024-09-22.
Figure 8: Delta-days prediction absolute-error quantiles by true first-seen month. The dashed line marks the month containing the training maximum first-seen date, 2024-09-22.
Split
Label
N
Obvious-bad N
Rate
train
benign
1,170,000
282,924
0.2418
train
malicious
1,170,000
531,602
0.4544
test
benign
270,000
88,033
0.3260
test
malicious
270,000
129,198
0.4785
Table 7: Overall obvious-bad timestamp rates by split and label.
Split
Label
Zero
After first seen
Before 1995
Repeated
train
benign
0.2122
0.4865
0.3499
0.4277
train
malicious
0.1189
0.1419
0.3827
0.8604
test
benign
0.4226
0.4140
0.5131
0.5149
test
malicious
0.2246
0.1559
0.4693
0.8753
Table 8: Non-exclusive obvious-bad type rates among PE/.NET rows marked obvious-bad in each split-label group. The repeated column counts exact COFF timestamps observed at least 1,000 times in the compact PE/.NET timestamp corpus.
Table 9: Obvious-bad timestamp classifier performance. The model is a single XGBoost binary classifier trained on benign and malicious train rows.
Figure 9: Feature-family TreeSHAP differences for the obvious-bad timestamp classifier on the regular test split. Positive values indicate larger mean signed contribution for malicious rows; negative values indicate larger mean signed contribution for benign rows.
Feature family
Benign signed
Malicious signed
Δ
header
-0.6775
-0.0152
0.6623
section
-0.4576
0.1136
0.5712
imports
-0.3140
0.1834
0.4973
Rich header
-0.3393
-0.0810
0.2583
strings
-0.1188
0.0476
0.1664
authenticode
0.0649
-0.0986
-0.1634
Table 10: Regular-test feature-family TreeSHAP summaries for the obvious-bad timestamp classifier. Δ is malicious minus benign mean signed SHAP.
Figure 10: Regular-test feature-level SHAP differences for the obvious-bad timestamp classifier, sorted by malicious-minus-benign mean signed SHAP, with ranks 401–2199 omitted from the x-axis display. Values above zero indicate features with larger obvious-bad log-odds contribution for malicious rows; values below zero indicate larger contribution for benign rows.
Rank
Feature
Column
Benign
Malicious
Δ
1
Authenticode signing-time minus COFF time
2478
0.2658
-0.1728
-0.4386
2
Debug directory size
2416
0.3000
-0.0955
-0.3956
3
Optional header uninitialized-data size
714
-0.0907
0.1665
0.2572
4
Authenticode latest signing time
2477
-0.1541
0.0778
0.2319
5
PE warning: suspicious section flags
2546
-0.0641
0.0710
0.1351
6
Optional header major linker version
704
-0.2105
-0.0776
0.1329
Table 11: Largest regular-test individual-feature signed TreeSHAP differences for the obvious-bad timestamp classifier.
Figure 11: Range of Spearman correlations between original EMBER2024 malware probabilities and signed timestamp-artifact family SHAP contributions on the regular test split. Each box summarizes nine values: three original malware models crossed with the all, benign, and malicious row groups.