Welcome to conference reviewing. Your situation is unfortunately extremely common. If the number of submissions in area X is less than the number of reviewers with interest and expertise in area X, then inevitably some of those reviewers will be assigned papers in area Y. Symmetrically, if the number of submissions in X is more than the number of experts in X, then some of those papers must be assigned to reviewers who'd prefer Z. Ideally, most papers are reviewed by experts and each reviewer is assigned mostly papers in their areas of expertise, but sometimes the matching system breaks down.
By default, you must review the papers that you were assigned. But there are a couple of ways to address mismatches.
If your assignment looks like a complete mismatch, you should quickly contact the PC chair and express your concerns, and in particular to verify that they received your preference email. If you're early enough in the process, the chair may be able and willing to swap assignments with other reviewers with complementary mismatches. There might also be a mechanism to swap assignments directly with other committee members, but you should always get the PC chair's permission before actually swapping.
Most computer science conference review forms ask for your confidence in your review; fill in this field honestly. If you really don't have the expertise to understand the paper or make a recommendation, your review should read "I apparently do not have the expertise to understand this paper or make a recommendation." That's a damning review of both the paper and the assignment process, but that's okay; your job as a reviewer is to be respectful but brutally honest.
On the other hand, as a matter of simple self-preservation, you should not review papers that you were not assigned. Reviewing well is extremely hard and time-consuming, and you have many other demands on your time. (And reviewing badly should just not be done at all.) Generally speaking, doing extra unsolicited reviewing will not earn you additional credit or brownie points.
There are a few exceptions. If you see a paper that you plan to read in detail anyway, despite getting no credit for doing so, you might as well write a review. And if the initial reviews of a paper disagree, the PC chair may seek out additional input or even full reviews; if you're asked about a paper in your area of expertise late in the review cycle, you should help. (If you're asked about a paper outside your area of expertise in later rounds, you should just say no.)
