A lot of people have already shown some of the numbers problems with balancing INT and Def, but there is a more fundamental gameplay reason why this should not be done:
The critical balance between Might and Magic.
A hero who puts all his effort in to upping his might should have a vulnerability to magic, an area he has neglected. Likewise, a hero who has put all his effort in to his spellcasting is, well, extremely squishy. 
This is a necessary balancing mechanism to make each choice meaningful and with consequences. There shouldn't be just one right approach.
If a might focused player can trick out all the weapons and armor, and have that same Def which works against other might heroes also nullify magic heroes spells, it breaks that balance. One solution, Big Shiny Armor, solves both his problems. Meanwhile the Magic hero, with his cloth armor and spiffy staff and rings is still 
.
When it comes to the second problem of what to roll against to block spells, the right answer is...you don't need to!
Unlike physical damage, spells are already being balanced by a number of other limitations that, when spells can also fail, makes them too weak and unreliable an option. These limitations are:
1. Mana. Mana is in short supply. If a spell is too powerfull, the mana cost can always be raised. Keep in mind that a spell caster without magic is dead, whereas a might hero can always keep attacking turn after turn meaning misses and bad rolls are less devastating. Mana also takes many turns to regenerate further increasing the cost of using it.
2. Magic users are extremely scarce in this game: Hundreds of turns may pass with only 1 person in your empire (your sovereign) able to cast. This makes magic even more critical that it work reliably, because if your sovereign is a mage and his spells fail to work in a battle it can be game over.
3. The only ways to up damage on spells are incredibly difficult and time consuming. Leveling and upping int and gaining shards of the right type being the major ways. On the other hand, weapons for upping attack are really easy to come by. For this reason, the damage curve of a spellcaster as opposed to a warrior is easier to control.
Edit: A spell Resistance attribute that blocks some damage from spells could be good for additional tactics and strategy, i.e SR 10% would block 10% of spell damage. But all units do not need it to start. Certain creatures can have spell resistance, and certain gear as well in exchange for lower defense. Also spells can give resistance, etc., but a standard starting value that can block all spell damage on certain rolls is a bad idea.
For the issue of shards, the benefit for acquiring them needs to be great enough to be worth the expense, because unlike might improvements, shards can be attacked, causing a mage to lose all their power. It costs time, money, and effort getting the land to control shards and that needs a strong enough effect to make attacking them a real strategic option in a game. If they are two weak, no one will care about obtaining or attacking them, which is a waste of a very interesting element of the game.
The best approach would be to use a curved power scale, with bigger power jumps for the first shard or two and lesser benefits as more are acquired. A good curve should avoid brokenness of acquiring lots of shards that might happen later in the game as adventure tech opens up. Something like:
First shard 2x
Second shard 3x
Third Shard 3.5x
Fourth Shard 3.75x
and so on. I'm sure it will need tweaking but the important thing is the shape of the curve.
Also with shards, it's very likely magic will become more differentiated later on. I also see a good chance of players needing to choose spell schools more carefully after further patches; It's too easy to acquire all schools right now making all heroes the same. If more choice and diversification does come in later patches, finding and capturing the right shards may become even harder to do.
Anyway, please keep these ideas in mind as you move forward with your changes. Magic needs a lot of love, but when it starts working it will be good times. 
