🛡️ Sentinel: [MEDIUM] 프롬프트 검증 시 발생하는 정수 오버플로우 취약점 수정

.jules/sentinel.md

@@ -2,3 +2,7 @@
 **Vulnerability:** Unvalidated inputs passed to `if()` statements can cause process crashes (`condition has length > 1`) or unexpected coercion vulnerabilities.
 **Learning:** In R, optional boolean parameters that default to `NULL` should be validated using explicit runtime type validation (e.g., `if (!is.null(flag) && (!is.logical(flag) || length(flag) != 1 || is.na(flag)))`).
 **Prevention:** Always implement explicit runtime type validation for optional boolean parameters.
+## 2024-09-27 - Fix unconstrained integer validation
+**Vulnerability:** Unconstrained regex matching `grepl("^[0-9]+$", n)` allows excessively large numbers to be converted by `as.integer(n)`, resulting in integer overflow (`NA` warning) which crashes downstream conditional checks expecting a boolean or integer outcome.
+**Learning:** In R, validating input bounds is important especially for menu selections. Overly broad regexes leave the type coercion vulnerable to overflow logic errors.
+**Prevention:** Rather than regex checks and coercion, validate the raw string explicitly against expected enum values (e.g., `n == "1" || n == "2"`) to safely guard internal logic flow.

DESCRIPTION

@@ -10,7 +10,7 @@ Description: Automates fixed item parameter linking for test linking under
     the item response theory paradigm using mirt package estimates.
 License: GPL-3 | file LICENSE
 Imports: mirt, methods
-Suggests: testthat (>= 3.0.0)
+Suggests: testthat (>= 3.0.0), mockery
 Encoding: UTF-8
 Config/testthat/edition: 3
 Config/roxygen2/version: 8.0.0

R/aFIPC.R

@@ -141,7 +141,7 @@ autoFIPC <-
       }
       for (attempt in seq_len(3)) {
         n <- readline(prompt = "Is it correct? (1: Yes 2: No) : ")
-        if (grepl("^[0-9]+$", n)) {
+        if (n == "1" || n == "2") {
           return(as.integer(n))
         }
       }
@@ -171,7 +171,7 @@ autoFIPC <-
               readline(
                 prompt = "Do you want to use default BILOG-MG priors for oldform Data? (1: Yes 2: No) : "
               )
-            if (grepl("^[0-9]+$", n)) {
+            if (n == "1" || n == "2") {
               return(as.integer(n))
             }
           }
@@ -390,7 +390,7 @@ autoFIPC <-
               readline(
                 prompt = "Do you want to use default BILOG-MG priors for newform Data? (1: Yes 2: No) : "
               )
-            if (grepl("^[0-9]+$", n)) {
+            if (n == "1" || n == "2") {
               return(as.integer(n))
             }
           }

tests/testthat/test-sentinel-integer-validation.R

@@ -0,0 +1,27 @@
+test_that("autoFIPC integer overflow in readline input parsing is prevented", {
+  # Provide small dataset
+  dummy_data <- data.frame(
+    Item1 = c(1, 0, 1, 0, 1, 1),
+    Item2 = c(0, 1, 0, 1, 0, 0),
+    Item3 = c(0, 0, 1, 1, 0, 0),
+    Item4 = c(1, 1, 0, 0, 1, 1)
+  )
+
+  # Mock interactive to return TRUE
+  mockery::stub(aFIPC::autoFIPC, "interactive", function() TRUE)
+
+  # Mock readline to return a huge number first, then a valid "1"
+  mockery::stub(aFIPC::autoFIPC, "readline", mockery::mock("100000000000000", "1", cycle = TRUE))
+
+  # As we fixed the bug, it won't crash with "missing value where TRUE/FALSE needed" NA coercion error.
+  # We use try() to catch whatever it does down the line (which is estimation failing).
+  expect_error(
+    aFIPC::autoFIPC(
+      newformXData = dummy_data,
+      oldformYData = dummy_data,
+      newformCommonItemNames = "Item1",
+      oldformCommonItemNames = "Item1",
+      confirmCommonItems = NULL
+    ), regexp = "completely failed|could not estimate"
+  )
+})