diff --git a/include/exiv2/value.hpp b/include/exiv2/value.hpp
index f6b0894b32..cd9c2e0a48 100644
--- a/include/exiv2/value.hpp
+++ b/include/exiv2/value.hpp
@@ -1628,7 +1628,7 @@ namespace Exiv2 {
     std::string ValueType<T>::toString(long n) const
     {
         ok_ = true;
-        return Exiv2::toString<T>(value_[n]);
+        return Exiv2::toString<T>(value_.at(n));
     }
 
     // Default implementation
@@ -1636,69 +1636,69 @@ namespace Exiv2 {
     long ValueType<T>::toLong(long n) const
     {
         ok_ = true;
-        return static_cast<long>(value_[n]);
+        return static_cast<long>(value_.at(n));
     }
-// #55 crash when value_[n].first == LONG_MIN
+// #55 crash when value_.at(n).first == LONG_MIN
 #define LARGE_INT 1000000
     // Specialization for rational
     template<>
     inline long ValueType<Rational>::toLong(long n) const
     {
-        ok_ = (value_[n].second != 0 && INT_MIN < value_[n].first && value_[n].first < INT_MAX );
+        ok_ = (value_.at(n).second != 0 && INT_MIN < value_.at(n).first && value_.at(n).first < INT_MAX );
         if (!ok_) return 0;
-        return value_[n].first / value_[n].second;
+        return value_.at(n).first / value_.at(n).second;
     }
     // Specialization for unsigned rational
     template<>
     inline long ValueType<URational>::toLong(long n) const
     {
-        ok_ = (value_[n].second != 0 && value_[n].first < LARGE_INT);
+        ok_ = (value_.at(n).second != 0 && value_.at(n).first < LARGE_INT);
         if (!ok_) return 0;
-        return value_[n].first / value_[n].second;
+        return value_.at(n).first / value_.at(n).second;
     }
     // Default implementation
     template<typename T>
     float ValueType<T>::toFloat(long n) const
     {
         ok_ = true;
-        return static_cast<float>(value_[n]);
+        return static_cast<float>(value_.at(n));
     }
     // Specialization for rational
     template<>
     inline float ValueType<Rational>::toFloat(long n) const
     {
-        ok_ = (value_[n].second != 0);
+        ok_ = (value_.at(n).second != 0);
         if (!ok_) return 0.0f;
-        return static_cast<float>(value_[n].first) / value_[n].second;
+        return static_cast<float>(value_.at(n).first) / value_.at(n).second;
     }
     // Specialization for unsigned rational
     template<>
     inline float ValueType<URational>::toFloat(long n) const
     {
-        ok_ = (value_[n].second != 0);
+        ok_ = (value_.at(n).second != 0);
         if (!ok_) return 0.0f;
-        return static_cast<float>(value_[n].first) / value_[n].second;
+        return static_cast<float>(value_.at(n).first) / value_.at(n).second;
     }
     // Default implementation
     template<typename T>
     Rational ValueType<T>::toRational(long n) const
     {
         ok_ = true;
-        return Rational(value_[n], 1);
+        return Rational(value_.at(n), 1);
     }
     // Specialization for rational
     template<>
     inline Rational ValueType<Rational>::toRational(long n) const
     {
         ok_ = true;
-        return Rational(value_[n].first, value_[n].second);
+        return Rational(value_.at(n).first, value_.at(n).second);
     }
     // Specialization for unsigned rational
     template<>
     inline Rational ValueType<URational>::toRational(long n) const
     {
         ok_ = true;
-        return Rational(value_[n].first, value_[n].second);
+        return Rational(value_.at(n).first, value_.at(n).second);
     }
     // Specialization for float.
     template<>
@@ -1706,7 +1706,7 @@ namespace Exiv2 {
     {
         ok_ = true;
         // Warning: This is a very simple conversion, see floatToRationalCast()
-        return floatToRationalCast(value_[n]);
+        return floatToRationalCast(value_.at(n));
     }
     // Specialization for double.
     template<>
@@ -1714,7 +1714,7 @@ namespace Exiv2 {
     {
         ok_ = true;
         // Warning: This is a very simple conversion, see floatToRationalCast()
-        return floatToRationalCast(static_cast<float>(value_[n]));
+        return floatToRationalCast(static_cast<float>(value_.at(n)));
     }
 
     template<typename T>
diff --git a/src/crwimage_int.cpp b/src/crwimage_int.cpp
index 7b958c26f1..6e89027556 100644
--- a/src/crwimage_int.cpp
+++ b/src/crwimage_int.cpp
@@ -929,7 +929,7 @@ namespace Exiv2 {
         assert(pCrwMapping != 0);
         ULongValue v;
         v.read(ciffComponent.pData(), 8, byteOrder);
-        time_t t = v.value_[0];
+        time_t t = v.value_.at(0);
         struct tm* tm = std::localtime(&t);
         if (tm) {
             const size_t m = 20;
diff --git a/src/exif.cpp b/src/exif.cpp
index 50ecfa203f..263ab23dd3 100644
--- a/src/exif.cpp
+++ b/src/exif.cpp
@@ -230,7 +230,22 @@ namespace Exiv2 {
               fct=NULL;
             }
         }
-        if ( fct ) fct(os, value(), pMetadata);
+        if ( fct ) {
+          // https://github.com/Exiv2/exiv2/issues/1706
+          // Sometimes the type of the value doesn't match what the
+          // print function expects. (The expected types are stored
+          // in the TagInfo tables, but they are not enforced when the
+          // metadata is parsed.) These type mismatches can sometimes
+          // cause a std::out_of_range exception to be thrown.
+          try {
+            fct(os, value(), pMetadata);
+          } catch (std::out_of_range&) {
+            os << "Bad value";
+#ifdef EXIV2_DEBUG_MESSAGES
+            std::cerr << "Caught std::out_of_range exception in Exifdatum::write().\n";
+#endif
+          }
+        }
         return os;
     }
 
diff --git a/src/value.cpp b/src/value.cpp
index 7588869da2..64e4a9f863 100644
--- a/src/value.cpp
+++ b/src/value.cpp
@@ -251,7 +251,7 @@ namespace Exiv2 {
     {
         std::vector<byte>::size_type end = value_.size();
         for (std::vector<byte>::size_type i = 0; i != end; ++i) {
-            os << static_cast<int>(value_[i]);
+            os << static_cast<int>(value_.at(i));
             if (i < end - 1) os << " ";
         }
         return os;
@@ -260,7 +260,7 @@ namespace Exiv2 {
     std::string DataValue::toString(long n) const
     {
         std::ostringstream os;
-        os << static_cast<int>(value_[n]);
+        os << static_cast<int>(value_.at(n));
         ok_ = !os.fail();
         return os.str();
     }
@@ -268,19 +268,19 @@ namespace Exiv2 {
     long DataValue::toLong(long n) const
     {
         ok_ = true;
-        return value_[n];
+        return value_.at(n);
     }
 
     float DataValue::toFloat(long n) const
     {
         ok_ = true;
-        return value_[n];
+        return value_.at(n);
     }
 
     Rational DataValue::toRational(long n) const
     {
         ok_ = true;
-        return Rational(value_[n], 1);
+        return Rational(value_.at(n), 1);
     }
 
     StringValueBase::StringValueBase(TypeId typeId)
@@ -352,19 +352,19 @@ namespace Exiv2 {
     long StringValueBase::toLong(long n) const
     {
         ok_ = true;
-        return value_[n];
+        return value_.at(n);
     }
 
     float StringValueBase::toFloat(long n) const
     {
         ok_ = true;
-        return value_[n];
+        return value_.at(n);
     }
 
     Rational StringValueBase::toRational(long n) const
     {
         ok_ = true;
-        return Rational(value_[n], 1);
+        return Rational(value_.at(n), 1);
     }
 
     StringValue::StringValue()
@@ -404,7 +404,7 @@ namespace Exiv2 {
     {
         value_ = buf;
         // ensure count>0 and nul terminated # https://github.com/Exiv2/exiv2/issues/1484
-        if (value_.size() == 0 || value_[value_.size()-1] != '\0') value_ += '\0';
+        if (value_.size() == 0 || value_.at(value_.size()-1) != '\0') value_ += '\0';
         return 0;
     }
 
@@ -815,22 +815,22 @@ namespace Exiv2 {
     std::string XmpArrayValue::toString(long n) const
     {
         ok_ = true;
-        return value_[n];
+        return value_.at(n);
     }
 
     long XmpArrayValue::toLong(long n) const
     {
-        return parseLong(value_[n], ok_);
+        return parseLong(value_.at(n), ok_);
     }
 
     float XmpArrayValue::toFloat(long n) const
     {
-        return parseFloat(value_[n], ok_);
+        return parseFloat(value_.at(n), ok_);
     }
 
     Rational XmpArrayValue::toRational(long n) const
     {
-        return parseRational(value_[n], ok_);
+        return parseRational(value_.at(n), ok_);
     }
 
     XmpArrayValue* XmpArrayValue::clone_() const
diff --git a/test/data/issue_1706_poc.exv b/test/data/issue_1706_poc.exv
new file mode 100644
index 0000000000..375e4fa7cc
Binary files /dev/null and b/test/data/issue_1706_poc.exv differ
diff --git a/tests/bugfixes/github/test_issue_1706.py b/tests/bugfixes/github/test_issue_1706.py
new file mode 100644
index 0000000000..ef31c9bcd8
--- /dev/null
+++ b/tests/bugfixes/github/test_issue_1706.py
@@ -0,0 +1,24 @@
+# -*- coding: utf-8 -*-
+
+from system_tests import CaseMeta, path
+
+
+class InvalidDateXMP(metaclass=CaseMeta):
+    """
+    Regression test for the bug described in:
+    https://github.com/Exiv2/exiv2/issues/1706
+    """
+    url = "https://github.com/Exiv2/exiv2/issues/1706"
+
+    filename = path("$data_path/issue_1706_poc.exv")
+    commands = ["$exiv2 -PE $filename"]
+
+    stderr = [
+"""Error: Directory Photo with 65280 entries considered invalid; not read.
+"""
+]
+    retval = [0]
+
+    def compare_stdout(self, i, command, got_stdout, expected_stdout):
+        # Check that it printed "Bad value" for the date.
+        self.assertRegex(got_stdout, "Exif.PentaxDng.Date\\s+Long\\s+1\\s+Bad value")