Is it worth making?[/quote]
If you try to make your code reusable then even if your project doesn't get completed, you'll have a experience and a code base to work with on your next project rather than starting from scratch. So I say, yes, it's worth making.

If you would like to play a game like this?[/quote]

Over all, the idea sounds interesting and has potential. As for whether I'd play it had I just come across it on some website somewhere... well I might, depending on my mood and your ability to stand out from whatever other games your title sits next to (marketing is a whole different thing than programming). There's a lot of 'em after all and usually my decision to try a game is quick and impulsive and not the result of carefully reading details about the game. For my reply here to be a little more useful than that though, I read the feature list a bit more carefully than I probably would've a game description on a website. The things that seemed to draw me towards wanting to play were the things you had actually numbered 1-3 (also the random city levels). If you had lead with the descriptions you had that go with these points, I think my attention and interest would've been more focused.

If there's anything I'd suggest changing, it'd be to drop the leveling & perks system. I don't think it's all that necessary. You have the ability to set up defenses, the effectiveness of which will be due to (I'd expect) what the player learns to be more effective which sounds more fun to me than a nebulous stat in the background. You have weapons that, I presume will provide various degrees of effectiveness. What would increasing in levels really give you? Given that you want bullets to be used up really fast (presumably to maintain a tense atmosphere), would finding 5 instead of 2 really be worth creating a leveling system for, especially when the bazooka isn't far from reach? And if the player reaches a high level where the player feels like he is extremely powerful, how do you maintain the sense of dread of running out of bullets? A sense of progression is of course important to include in a game though so maybe rather than leveling up the player give him the ability to upgrade equipment. Since it's already established equipment runs out, the fear of running out of resources is ever present, even if those resources are a little more effective.

As for the Progression vs. Survival, I think Call of Duty: Zombie did it well. When you killed the zombie, you earned money, and through that money you were able to buy more guns and ammo. However, after certain amount of upgrades in your arsenal, there was a limit (usually a futuristic gun) to this, and it basically came down to the skill level.

Same can be achieved here. Make it hard to level up. But when the player does level up (stat wise, or equipment wise), reward them well, make the game easy for a night or two, giving the player a sense of achievement. However, the difficulty eventually catches up, and the player is forced to level up again, or die trying. And after few of these leveling(may be cap the level at 10) there's no more real progression and it all comes down to the player's skill to survive as long as possible, and the choices he made while leveling up (where he made his fortress, what kind of stat/skill/equipment he decided to get upon leveling up, etc.) At this point, the way player will feel progression is through the learning how to survive as long as possible, and playing again and again to somehow survive longer and longer nights.

You have to make sure that the player CAN theoretically survive forever though. Make the food/battery replenish-able, but hard to obtain, meanwhile zombies gain in number and strength linearly every night. Important here is that difficulty has to be increased linearly. Players should never have to say it's easy to get through 49th night but it's impossible to get through 50th night. "Going through 49th night was HARD, but man, that 50th level is even HARDER and I ended up dying at the very end. But with enough practice, I SHOULD BE able to get to the 51th night!" *Restarts the game*
For me, this is the ideal approach to any endless survival game.

In this type of game, I believe luck plays a big role, that it should matter in a way that will be fun for players, but players should not be relying on luck to progress.
In another word, luck should always REWARD player in addition to normal game play, rather than punishing player for not being lucky.

Good game design is this: You know you will be awarded 5 batteries if you go through this dungeon, aka subway. At the end of the hard boss fight to get that 5 batteries, you get lucky and find an additional battery, giving you 6. Or you stumble upon a supply of food in addition to that 5 batteries. Players are rewarded for being lucky, but if they weren't lucky, they still got what they wanted to achieve, which is that pack of 5 batteries. And when this happens, you have to let the player know, by saying: OOOOHH LOOK WHAT I FOUND! AN EXTRA PACK OF BATTERY/FOOD! This kind of luck makes the game fun, as long as you can balance out the significance vs frequency well (meaning how much reward will you be giving to the player and how that relates to how frequent you'll be "lucky" and be rewarded)

Bad game design is this: making the loot randomized, so player does not know what the reward is. He is in need of 5 batteries. After the hard boss fight, he is given 10 packs of food, but no battery packs. He now has 10 extra packs of food, but he dies because he was not lucky enough to get battery packs which he needs, instead of unnecessary food packs.

programming-wise, since this is your first attempt at the game, you should focus on the main feature of the game: survival.
Forget the leveling, forget the dynamic environment, forget the multiple houses, and forget the multiple guns.
Keep the map static (make the map beforehand on paper or something and just hardcode in the data). In the map, make sure you have the keypoints (dungeons, etc.) at the reasonable places. Give the player 1 house, 1 gun, few battery packs to be used in home, few ammo, and few food in the beginning. During the day all player can do is just walk around and explore the map, as all the other features (looking for a possible house to live in, looking for other guns, etc) can be implemented later on.
Focus on the night part the most. Make sure zombies and boss AI is good and challenging. Make sure you code it so that their difficulty can be increased linearly, and infinitely.

And when you have those key features down and playable, you can add in additional features such as dynamic map creation, leveling, multiple houses, more equipments, random events, etc. After you implement those features, balancing should be relatively easy since it'll most likely be just changing the numbers around on the zombie's stat.

Edit:
Leveling:
As for the leveling, if I were to develop this game and implement stats, I would not add in the usual strength, agility, intellect stuff in there. Instead, I would give more discrete terms:
-Armor (either directly or indirectly. Directly meaning it would act sort of like HP, that if you increase this, you automatically get increase in how much damage you can take before you die. Indirectly meaning you must raise this to certain amount if you want to equip this armor.)
-Weapon efficiency (I'm assuming you won't be punching the zombies. With this, you increase the damage you do with weapons.)
-Stamina (How long you can survive without eating. Or how much stuff you can carry. Or both.)
-Architect (How much damage/armor the fortification your building has. Also this might decrease the time it takes to repair such buildings.)

Another good way to add in leveling is tech tree. You can decide to go for weapon route, and upon leveling, you can unlock certain weapon. Or you can decide to unlock certain building instead of weapon.

The point is, with some good brainstorming, there are so many ways you can implement leveling in infinite survival game.

Graphics:
If you want to add in "height" factor, I think isometric works the best. Although a bit more work on sprites/engine, I think it just allows a lot more interactive feeling with the environment, but this is a subjective matter

Multiplayer:
A nice coop mode and/or competition mode would make this game really more appealing if you were to publish it. However, I would say worry about this at the very end when you have finished making a playable singleplayer game. Just something to think about.

Hope this helped and good luck!