We examine bounded retained history through a frozen predictive pilot and a subsequent allocation study on a curated UNSW-NB15 flow mirror. Both use an artificial release-order replay based on raw ltime ticks, with equal-tick records querying a shared prior-group state. In the pilot, the model with added retained relational features had lower reconnaissance recall than the current-flow-plus-counters baseline (43.66% versus 47.84%) and lower benign false-positive rate (3.63% versus 6.68%) at separately frozen operating thresholds, together with lower AUROC. This mixed result did not establish a predictive advantage. A development-only audit subsequently found that capacity eviction dominated expiration in the 256-record, 300-tick FIFO state, with a median completed record lifetime of 11 ticks. A protocol specified before its single execution, but after the pilot and post-hoc mechanism audit, compared five label-blind retention arms on the already-used development trace. Its primary equal-source endpoint averaged retained/reference ratios for service-expanded source out-degree and source-destination service diversity. Source-balanced retention at 256 records exceeded FIFO-256 by a pooled endpoint difference of 0.440738 and met the prespecified materiality and subblock rule. Crucially, the event-weighted ratios for both quantities decreased relative to FIFO-256: the result is specific to the equal-source weighting objective and does not establish uniformly improved retention. The study supplies neither evidence of detection improvement nor an explanation of the pilot result. Reused data, concentrated reconnaissance support, proxy timing, and the absence of a qualified independent corpus bound the interpretation.
Figures & tables
Frozen pilot role
Rows
Reconnaissance rows
Training
775,008
1,361
Tuning
104,620
215
Calibration
143,569
164
Final test
1,036,218
11,617
Frozen pilot role support
Model
Recon. recall
Benign FPR
Other-attack misattribution
AUROC
Average precision
Brier score
Log loss
B0
0.478092
0.063653
0.303534
0.734002
0.137139
0.014194
0.156281
B1
0.478437
0.066787
0.303317
0.735021
0.123377
0.013746
0.189880
B2
0.436602
0.036291
0.231722
0.678465
0.120835
0.014084
0.125154
Calibration-prior constant
0
0
0
0.500000
0.011211
0.011187
0.077081
Table 1: Frozen calibrated pilot results. Each stream uses its own previously frozen operating threshold.
Arm
Record cap
Eviction policy
FIFO-64
64
Global insertion-order FIFO
FIFO-256
256
Global insertion-order FIFO; primary baseline
FIFO-1024
1024
Global insertion-order FIFO
SRC-256
256
Evict the oldest record of the currently most occupied retained source
DIV-256
256
Source-balanced victim-source selection, with preference for evicting a duplicate relation
Prespecified retention arms
Arm
Pooled M
Equal-source Os
Equal-source Vsd
FIFO-64
0.091471
0.073888
0.109055
FIFO-256
0.183633
0.168766
0.198501
FIFO-1024
0.319540
0.318420
0.320660
SRC-256
0.624371
0.621398
0.627345
DIV-256
0.645849
0.646344
0.645355
Table 2: Pooled retention endpoints for all prespecified arms. Values are rounded to six decimals from summary.json ; Os and Vsd columns are equal-source mean ratios. Arithmetic and the decision rule use the stored machine-precision values, not the rounded display.
Aggregation
FIFO-256 M
SRC-256 M
SRC-256 minus FIFO-256
Sources Os / Vsd
Pooled
0.183633
0.624371
0.440738
35 / 34
S1
0.177710
0.622169
0.444459
35 / 34
S2
0.083780
0.619761
0.535980
25 / 25
S3
0.136080
0.622703
0.486623
25 / 25
Table 3: Prespecified primary contrast and fixed subblocks. Source counts are eligible contributing sources for Os / Vsd , not scanner counts.
Quantity
FIFO-256
SRC-256
Os
0.066887
0.061185
Vsd
0.072118
0.065100
Table 4: Pooled event-weighted ratios for the primary arms. These are averages over eligible queries, using the same reference-positive eligibility rule.
Network intrusion detection identifies malicious or abnormal network activity to help protect computers, connected devices, and critical infrastructure.